# Token Perks — full text
Research snapshot Sep 6–7 2026 — re-verify at official terms before paying.
Methodology v2 (2026-09-07): blended $/M = (3 x input + 1 x output) / 4 at each route's current published price (list, or launch-promo price while a promo runs — promo-priced rows caption their list-price blend); official sources only; uncertainty labels on every row; benchmarks quoted per datum with attribution, never republished as tables or feeds.
## Kimi K3 core membership [active]
Provider: Moonshot AI | Plan: Kimi K3 Membership — Moderato / Allegretto / Allegro / Vivace
Price now: $19–$199/mo by tier (Moderato $19, Allegretto $39, Allegro $99, Vivace $199)
Renewal: Renews monthly at the same tier price; annual billing works out to about $15 / $31 / $79 / $159 per month of effective cost.
Verified: 2026-09-06
Verdict: Pick Kimi K3 membership if you run steady agentic coding volume and want one predictable bill. At Allegretto ($39/mo, about $31/mo effective annually) it beats pay-as-you-go once you clear roughly 40–50 heavy tasks a month — below that, PAYG or a free route costs less. The constraint: a single shared credit pool plus 5-hour and weekly controls means marathon sessions can stall mid-sprint.
Access: Subscribe to Kimi K3 membership on the official site, then point coding tools at the coding endpoint (https://api.kimi.com/coding/) — not a PAYG API key — so usage draws from the membership pool.
Caveats: One shared credit pool: membership usage draws from a single balance, so parallel projects and long agent runs drain it together | 5-hour and weekly controls: burst caps can throttle you mid-day even when the monthly balance looks healthy | Tier jump is steep: Allegretto ($39) to Allegro ($99) is a 2.5x step — usage between those tiers is priced comparatively poorly | Annual discount is upfront cash: the ~20% effective saving requires a full-year prepay with terms you must confirm first
Limits: k3-256k context with high thinking-on: long sessions consume materially more than short completions | Coding endpoint (https://api.kimi.com/coding/) vs PAYG key billing are different routes — pick the wrong one and the membership may not apply | Weekly controls reset on the provider's schedule, not yours — confirm the reset window before planning sprints | Fair-use style gating can apply to extreme patterns; the weekly controls are the mechanism to watch
Eligibility: Anyone who can create a Kimi account and add an accepted payment method | Coding-endpoint users should confirm their tooling targets https://api.kimi.com/coding/ rather than PAYG key billing | Annual effective pricing requires paying a full year upfront
Restrictions: One shared credit pool across native membership usage — heavy features draw from the same balance | 5-hour and weekly usage controls can gate burst usage even with credits remaining | Model served as k3-256k with high thinking-on behaviour; output style and verbosity affect consumption | Annual plan is upfront; confirm whether mid-cycle upgrades, downgrades, or refunds are offered
Think in effective cost per task: monthly price divided by tasks you actually complete. Allegretto at $39 over 120 tasks is about $0.33 per task — less than half the $0.80 illustrative PAYG reference for a 100k-token task.
The break-even against the $0.80/task reference is about 49 tasks a month on Allegretto ($39 ÷ $0.80) and about 24 tasks on Moderato ($19 ÷ $0.80). Below those lines, PAYG or a free tier is cheaper.
Annual billing drops Allegretto to about $31/mo effective (~$372 upfront), which moves break-even down to roughly 39 tasks a month — worth it only if your volume is stable year-round.
The shared pool has a throughput cost too: if 5-hour or weekly controls idle you for half a day each week, discount the membership's value by that lost throughput before comparing.
Q: What does Kimi K3 membership cost as of the Sep 2026 snapshot? A: Moderato $19, Allegretto $39, Allegro $99, and Vivace $199 per month. Annual billing lowers the effective monthly cost to about $15 / $31 / $79 / $159, paid upfront. Prices verified in the Sep 6 2026 snapshot — re-check the official page before paying.
Q: How does the shared credit pool restrict usage? A: Membership usage draws from one shared credit pool, and 5-hour plus weekly controls can throttle burst usage even when monthly balance remains. Spread heavy agent runs across the week instead of one marathon day.
Q: Coding endpoint or PAYG — which do I use? A: Use the coding endpoint at https://api.kimi.com/coding/ for membership billing. A PAYG API key bills separately, so pointing tools at the wrong route means the membership does not cover that usage.
Q: When is annual billing worth it? A: When your monthly volume is stable and clears break-even (about 39 tasks/mo on Allegretto annual vs ~49 on monthly, against the $0.80/task reference). If your usage is seasonal, stay monthly until the pattern is proven.
Q: Is API credit use cheaper than a $20 subscription, and at what usage? A: At the $0.80/task reference, $19 of pay-as-you-go buys about 24 tasks — below that, credits or a free route are cheaper; above it, the $19 Moderato month is cheaper. The routes do not mix: a PAYG key bills separately from the coding-endpoint membership, so check which one your tools actually hit.
Q: How current is this information? A: This is a research snapshot from Sep 6 2026, not a live checkout feed. Prices, pools, and controls can change — re-verify at the official terms before paying.
Evidence: [high] Tier prices ($19 / $39 / $99 / $199 monthly; ~$15 / $31 / $79 / $159 annual effective) and the coding endpoint URL come from the Sep 6 2026 research snapshot of official pages. (Source: Official Kimi coding page, captured Sep 6 2026) | [medium] Membership shares one credit pool with 5-hour and weekly controls — documented gating behaviour, exact thresholds may vary by account and tier. (Source: Official terms snapshot Sep 6 2026; thresholds not independently load-tested) | [low] Break-even figures are arithmetic on the $0.80/task illustrative reference, not measured per-user costs — your tokens-per-task will differ. (Source: Token Perks calculation; re-run with your own usage)
URL: https://token-perks.com/best/kimi-k3-core/
## OpenCode Zen Muse Spark 1.3 free promo [active]
Provider: Meta Muse Spark via OpenCode Zen | Plan: Muse Spark 1.3 Contributor Free — limited-time promo
Price now: $0 for input, cache, and output tokens during the promo window
Renewal: No renewal — this is a limited-time promo, not a plan. When it ends, usage moves to whatever paid terms apply, so confirm before building on it.
Expiry: Limited-time promo — end date confirmed only inside the official product; re-verify before relying on it.
Verified: 2026-09-07
Verdict: If you can see the exact string “Muse Spark 1.3 Contributor Free” under /connect → Zen → /models, use it at $0 for input, cache, and output while the promo lasts. In short: great for evaluation, prototyping, and light tasks today; a bad foundation for anything you cannot migrate in a day, because the promo ends on the provider's schedule, not yours.
Access: Inside the official product, navigate /connect → Zen → /models and select the exact string “Muse Spark 1.3 Contributor Free”. Do not guess at similarly named entries — only that exact string carries the $0 promo.
Caveats: Limited-time promo: $0 today can become metered tomorrow, with the end date visible only in-product | Exact-string requirement: only “Muse Spark 1.3 Contributor Free” is free — a near-identical paid entry may sit one row away | No price lock: there is nothing to renew and no grandfathering to assume | Route-dependent: the /connect → Zen → /models path is the documented access; if the UI changes, the promo may be unreachable
Limits: $0 covers input, cache, and output tokens under the promo — the full meter, not a slice | Throughput and concurrency limits are unpublished in the snapshot; expect free-tier shaping at peak | Promo may differ by account, region, or app version — a friend seeing it does not guarantee you see it | Anything stateful you build (fine-tunes, cached pipelines) should be portable before the window closes
Eligibility: Users who can reach the Zen section of the official product and see the promo model listed | Must select the exact string “Muse Spark 1.3 Contributor Free” — similarly named paid entries bill normally | Promo availability may vary by account, region, or app version
Restrictions: Limited-time promo — the $0 rate can end without notice outside the official window | Only the exact “Muse Spark 1.3 Contributor Free” string is $0; nearby model names may be paid | Reasonable-use gating may apply even at $0; it is free, not unlimited | No renewal or price lock — post-promo terms are whatever the provider publishes next
During the promo the effective cost per task is $0, which beats every paid route at any volume — including the $0.80/task PAYG reference and the $40/mo subscription (break-even: 50 tasks).
The accurate comparison is against the promo's expected lifespan. If you expect 200 tasks before the window closes, that is ~$160 of avoided PAYG spend — real money, but one-time.
Do not annualize free: a $39/mo membership decision should be made on post-promo prices, with the promo treated as a free trial of arbitrary length.
Migration cost is the hidden line item. Keep prompts, evals, and outputs provider-portable so the end of the promo costs you an afternoon, not a rewrite.
Q: Is Muse Spark 1.3 really free? A: The Contributor Free promo is $0 for input, cache, and output tokens while it lasts. It is a limited-time promo, not a free-forever plan — confirm the window in-product before depending on it.
Q: How do I find the free entry? A: Navigate /connect → Zen → /models inside the official product and select the exact string “Muse Spark 1.3 Contributor Free”. Similarly named entries may be paid, so match the string exactly.
Q: What happens when the promo ends? A: Usage moves to whatever paid terms the provider publishes — this page does not guess at them. Keep your work portable so the transition costs an afternoon, not a rewrite.
Q: Should I skip a paid plan and just use this? A: Use it for evaluation and portable work during the promo window, but make paid-plan decisions on post-promo prices. If your volume clears ~50 tasks/mo at $0.80/task, a $40-class subscription is the durable answer.
Q: How current is this information? A: Research snapshot Sep 6 2026. Promo availability changes fast — re-verify in the official product before paying or planning around it.
Evidence: [high] $0 input/cache/output and the /connect → Zen → /models access path with the exact promo string are stated on the official OpenCode Zen docs page; the in-product listing was captured Sep 6 2026. (Source: OpenCode Zen docs (opencode.ai/docs/zen/), accessed Sep 7 2026; in-product snapshot Sep 6 2026) | [medium] The promo is limited-time, but the end date is confirmed only inside the official product — this page states the limit without inventing a date. (Source: Snapshot Sep 6 2026; window not published as a fixed date) | [low] Avoided-spend figures use the illustrative $0.80/task reference, not the provider's post-promo price, which is unpublished. (Source: Token Perks calculation; illustrative only)
URL: https://token-perks.com/best/muse-spark-zen-free/
## NVIDIA Build Kimi K3 free [active]
Provider: NVIDIA Build | Plan: Kimi K3 free for development and prototyping
Price now: $0 for development and prototyping use within account limits
Renewal: No paid renewal in the snapshot — free dev/prototyping allocation governed by account-level limits, which vary. Confirm current quota in the official console.
Verified: 2026-09-06
Verdict: A $0 way to prototype on Kimi K3 with reasoning and tool calls intact. $0 for dev work within your account's limits, and the model behaviour you validate here — reasoning traces plus tool calls — carries over when you graduate to paid serving. The limiting factor is quota opacity: limits vary by account, so measure your actual ceiling on day one instead of trusting any number on this page.
Access: From the official NVIDIA Build catalog, find the Kimi K3 listing, confirm the free development/prototyping terms on your account, then integrate against the endpoint the console gives you — keeping reasoning and tool-call parameters enabled.
Caveats: Limits vary by account: two developers can see two different ceilings, and both can change | Dev/prototyping scope only: this is a workbench, not a production SLA | No published quota table in the snapshot — your real limit is whatever the console enforces today | Catalog rotation: free listings can narrow or move, stranding hardcoded endpoints
Limits: Reasoning and tool-call behaviour is preserved in the free route — that is the point of prototyping here | Rate limits and quotas are account-level and variable; test yours before demo day | Concurrency and context behaviour should be validated per account, not assumed from another team's numbers | Production traffic needs paid terms with explicit limits — confirm in the official console
Eligibility: Developers with an NVIDIA Build account in good standing | Development and prototyping workloads — not a production serving commitment | Account-level access to the Kimi K3 listing in the official catalog
Restrictions: Limits vary by account — rate limits, quotas, and availability differ and can change | Development/prototyping scope — production load belongs on paid terms | Catalog availability can shift without notice; listings rotate | Usage is subject to the platform's acceptable-use and rate-limit policies
Prototyping here costs $0 against the same illustrative $0.80/task PAYG reference — every 100 prototype tasks avoids ~$80 of paid inference while you validate reasoning and tool-call behaviour.
The value is information, not just savings: confirming that Kimi K3's reasoning traces drive your agent correctly is worth more than the avoided spend, because it de-risks the later paid commitment.
Budget the graduation: a prototype doing 120 tasks/mo that moves to a $40/mo subscription breaks even at 50 tasks against the $0.80 reference — plan that migration cost from day one.
Do not price production on free quota. Account-variable limits mean this page cannot promise a ceiling — your console's numbers are the only ones that count.
Q: Is Kimi K3 on NVIDIA Build really free? A: For development and prototyping within your account's limits, yes — $0 in the Sep 6 2026 snapshot. Limits vary by account, so confirm your quota in the official console.
Q: Do reasoning and tool calls work on the free route? A: Yes — reasoning and tool-call behaviour is preserved, which is what makes this route useful for validating agents before paying for serving.
Q: Can I run production on it? A: No. This is a development/prototyping allocation with variable, unpublished-per-account limits and no production SLA. Move serving traffic to paid terms.
Q: Why won't you quote the exact quota? A: Because limits vary by account and change over time — any single number here would mislead half our readers. Test your ceiling in the console on day one.
Q: Which AI coding tools are actually free with no credit card, and what are the limits? A: Two routes are free at the point of use in our Sep 6 snapshot: this NVIDIA Build K3 dev access and the Zen Muse Spark $0 promo. Whether signup requires a card varies by account and region, so we do not promise card-free onboarding. Limits differ: account-variable quotas here, a promo window that can end there.
Q: How current is this information? A: Research snapshot Sep 6 2026. Catalog listings and quotas rotate — re-verify in the official console before building plans on it.
Evidence: [high] Free dev/prototyping positioning with reasoning and tool calls preserved comes from the Sep 6 2026 research snapshot of the official listing. (Source: Official NVIDIA Build catalog listing, captured Sep 6 2026) | [medium] Limits vary by account — stated qualitatively because no single quota number applies to all readers. (Source: Snapshot Sep 6 2026; per-account enforcement observed) | [low] Avoided-spend figures use the illustrative $0.80/task reference; actual savings depend on your tokens-per-task and post-prototype serving price. (Source: Token Perks calculation; illustrative only)
URL: https://token-perks.com/best/nvidia-k3-free/
## GitHub Copilot Pro [active]
Provider: GitHub | Plan: Copilot Pro (individual)
Price now: $10/user/mo (monthly billing; annual price for Pro not published)
Renewal: Monthly per-user charge. No auto-renew terms page was found in this pass — the docs describe cancellation and proration but not renewal notice, so treat renewal mechanics as unverified.
Verified: 2026-09-07
Verdict: At $10/mo Copilot Pro is the cheapest tracked coding-tool subscription, and completions-only users effectively get unlimited service — the credit pool only matters if you lean on chat, agents, or premium models. Roughly $0.15 of metered usage is included daily on a 30-day month; a few Opus-class agent sessions can consume the pool, which is why the budget-and-alert mechanism exists. If your workflow is completion-first, Pro is strong value; if it is agent-first, budget for on-demand spend or price Pro+ ($39, $70/mo credits) alongside it.
Access: Subscribe from the Copilot plans page with a personal GitHub account, install the extension in VS Code/JetBrains, then check Settings → Copilot → Usage for your credit budget and alerts.
Caveats: The $10 sticker price buys $10 of base credits + $5 flex — not unlimited premium usage; chat/agent/model calls meter against it | Credits reset monthly with no rollover: a light January does not subsidize a heavy February | The plans page's static model table looks identical across tiers — the docs comparison table, not the marketing page, is the reliable tier mapping | Overage requires an explicitly set budget ('budgets do not stop license charges') and may still be capped mid-cycle
Limits: Included usage: $15/mo total credits on Pro (1,500 credits at $0.01) | Completions: unlimited on paid plans; Free is 2,000 completions + 50 chats/mo | Premium-model access: excluded at Pro tier per docs table; model options may vary by feature | Per-model token rates (what actually drains credits) live in the models-and-pricing reference, not the plans page | Exact daily caps beyond credits: not published
Eligibility: Any individual GitHub account with a supported payment method ($10/user/mo) | Free alternative exists: Copilot Free (2,000 completions + 50 chat requests/mo) and a Student plan for verified students at $0 | Teachers and maintainers of popular open-source projects: docs state access at no cost (eligibility mechanics beyond 'verified' not published) | Additional-credit purchase is unavailable for iOS/Android in-app subscriptions per the plans table
Restrictions: AI credits (plain-English: monthly usage dollars converted to points — 1 credit = $0.01) gate chat, agent mode, code review, cloud agent, CLI, and Apps; code completions and next-edit suggestions do not use credits and stay unlimited on paid plans | Pro carries $15/mo total credits = $10 base (1,000) + $5 flex (500); unused credits do not carry over — reset 00:00 UTC on the 1st of each month | Premium models (e.g. Claude Opus/Fable, GPT-5.5/5.6 Sol, GPT-6 Astra) are listed as Pro+/Max access in the docs tier table — Pro's own roster covers Sonnet 4.6/Haiku 4.5, GPT-5.4/5.6 Luna/Terra, Gemini 3.5 Flash, Grok 4.5/4.6, Kimi K2.7/K3, MAI-Code | Continuing past included credits requires setting a dollar budget; GitHub notes additional usage 'may be capped' | Code review also bills GitHub Actions minutes
Split the value: completions and next-edit suggestions are unmetered, so completion-only users get the full plan for $10 and never touch the credit pool.
Everything metered (chat, agents, review, CLI) draws credits at per-token model rates; $15/mo of credits is $0.50/day of model calls — one long Opus-class agent session can approach the whole base pool, which is why Pro's roster stops below the premium models.
Break-even framing against the $0.80/task illustrative reference: $10/mo covers roughly 12 metered tasks before on-demand is cheaper-or-dearer parity; below ~12 tasks/mo of chat/agent work, a budget-only setup on Free + occasional usage may cost less — but Free caps chats at 50/mo, so heavy users cross to Pro quickly.
Overage is opt-in: with no budget set you stop at the pool; with a budget you pay $0.01/credit in arrears. Model the pool as a ceiling you must consciously lift, not a soft cap.
Q: What does GitHub Copilot Pro cost as of the Sep 2026 snapshot? A: $10/user/mo, billed monthly. The plans page and billing docs both state it; no annual Pro price is published. Prices verified live 2026-09-07 — re-check before paying.
Q: What are AI credits and how many does Pro include? A: Credits are GitHub's metered usage unit at $0.01 each. Pro includes $15/mo total: a $10 base allotment (1,000 credits) plus a $5 flex allotment (500). Completions don't use credits; chat, agents, code review, cloud agents, CLI, and Apps do. Unused credits are forfeited at the monthly reset.
Q: What happens when Pro credits run out? A: Officially you can wait for the next cycle, switch to cheaper models, or set a dollar budget for additional usage billed in arrears at per-token model rates. GitHub notes additional usage may be capped, and budgets do not stop license charges.
Q: Can I use frontier premium models on Pro? A: Not according to the docs tier table, which lists Opus/Fable-class and GPT-5.5/5.6-Sol-class models under Pro+ and Max. Pro's own roster still includes Claude Sonnet 4.6/Haiku 4.5, GPT-5.4/5.6 Terra/Luna, Gemini 3.5 Flash, Grok 4.5/4.6, and Kimi K3. Model options may vary by feature.
Q: Is there a free path? A: Yes — Copilot Free (2,000 completions + 50 chat requests/mo, no credit card) and a Student plan at no cost for verified students; teachers and popular-OSS maintainers are also stated free. The Student plan is auto-model-selection only and excludes third-party agents.
Q: How current is this information? A: A live pass on 2026-09-07 across the plans page and three docs articles. Billing mechanics change often — re-verify at official docs before relying on budget behavior.
Evidence: [high] Pro $10/user/mo and the $15/$10+$5 credit split were read live on the official plans comparison page (2026-09-07), and the dollar and credit framings agree across plans page, billing docs, and usage-based-billing docs. (Source: github.com/features/copilot/plans + docs.github.com, accessed 2026-09-07) | [high] No-rollover, monthly reset at 00:00 UTC, budget mechanics (alerts at 75/90/100%, 'budgets do not stop license charges'), and possible capping of additional usage are stated in the official usage-based-billing doc. (Source: docs.github.com usage-based-billing-for-individuals, accessed 2026-09-07) | [medium] Tier-to-model mapping (premium models gated to Pro+/Max) comes from the docs plans table; the features page static table renders an identical roster across tiers, so the docs version is authoritative here. (Source: docs.github.com/en/copilot/get-started/plans, accessed 2026-09-07) | [low] Annual Pro pricing and auto-renewal notice terms were not found on any page fetched this pass — recorded as not published, not guessed. (Source: Absence across plans page + billing docs, 2026-09-07)
URL: https://token-perks.com/best/copilot-pro/
## ChatGPT Plus subscription [active]
Provider: OpenAI | Plan: ChatGPT Plus
Price now: $19.99/mo in the official US App Store IAP list; the Plus help article is excerpt-quoted at '$20/month, billed monthly'
Renewal: Help-center excerpts state Plus 'automatically renews at the standard ChatGPT Plus monthly rate unless canceled' and advise canceling at least 24 hours before renewal. No annual plan: the Plus article is excerpt-quoted saying annual billing or multi-month prepay is not supported.
Verified: 2026-09-07
Verdict: ChatGPT Plus remains the category-default $20 subscription: broader model and tool access than Free, Codex and Sora adjacent perks tied to ChatGPT plans, and a credit valve when you hit a cap. The honest caveat is evidentiary, not commercial — OpenAI's pages block automated verification, so this snapshot pairs a direct official App Store price with dated help-article excerpts and flags everything below list-price granularity as unconfirmed. Buy it for breadth at one flat fee; if you need published, stable caps (a fixed messages/day), no consumer chat sub we track publishes them — Plus is the most explicit that caps 'vary based on system conditions'.
Access: Subscribe from chatgpt.com (web billing, localized by country) or in the ChatGPT iOS/Android app (in-app purchase at store prices). After subscribing, watch Settings → Usage for weekly allowance banners and the credit/instant-reset controls.
Caveats: Caps vary by system conditions per OpenAI's own wording — the plan does not promise a fixed daily allowance | No annual discount exists at all for Plus (monthly only), unlike several rivals — steady users cannot prepay to cut cost | Which exact models Plus includes today is not reliably stated in anything retrievable — the marketing pages render client-side and help articles fragment by model launch | Credits and instant resets exist but their prices only surface in-product — the bill can meter past $20 in a heavy week
Limits: Message caps: per 5-hour windows and weekly allowances; the only recent numeric public figure found is the excerpted '3,000 msg/week GPT-5 Thinking' (May 2026 release note) | Codex, Work, uploads, image gen, voice each carry separate limits/reset periods per excerpts | Deeper usage tiers exist above Plus: help excerpt — 'Pro $100 unlocks 5x higher usage than Plus… Pro $200 unlocks 20x' | Free Go tier sits below at $8/mo in the same App Store list — Plus's relative headroom is not numerically published | Full current Plus feature list: not published in retrievable form this pass
Eligibility: Anyone who can create a ChatGPT account in a supported country and add payment (web card via Stripe-style localized billing, or in-app purchase) | Multi-currency localized pricing exists per the billing help article (excerpt) — displayed price varies by region | App-billed subscriptions (iOS/Android) carry the store's price, which may differ from web
Restrictions: Usage caps are model- and window-shaped: current framing is 5-hour rolling windows plus weekly allowances, per help release-note excerpts — OpenAI publishes no single consolidated Plus cap table | Excerpt from release notes: Plus had '3,000 messages/week with GPT-5 Thinking' and a 196k context for that model — account of record (May 2026); may have shifted | 'Plus subscriptions may include usage limits such as message caps, especially during high demand. These limits may vary based on system conditions.' (help excerpt) — caps are explicitly variable | Feature sub-limits are separate: 'ChatGPT file uploads, image generation, or voice have separate usage limits, reset periods, and banners' (Codex-tied article, excerpt) | Overage is credit-based: personal plans can buy flexible-usage credits to continue after hitting plan limits, and Plus/Pro can buy instant resets of weekly Work/Codex limits — credit prices: not published
Against the $0.80/task illustrative reference, $19.99/mo is parity at ~25 heavy tasks — and since Plus caps vary 'based on system conditions', model your month at the pessimistic end before trusting the flat fee.
No annual plan exists for Plus (per the help excerpt), so there is no prepay discount to compute — comparison against Claude Pro annual ($17/mo effective) is a real structural difference, not a pricing snapshot artifact.
Credits shift the plan toward hybrid pricing: the $20 buys a window-capped allowance, anything past it is metered at in-product credit prices we could not retrieve — treat marginal usage as unknown-cost, not free.
For code-agent workflows, Codex rides on ChatGPT plans with per-plan limits and buyable weekly resets — if you exceed Plus consistently, the excerpted Pro ladder (5x at $100) is the published next step.
Q: What does ChatGPT Plus cost as of the Sep 2026 snapshot? A: $19.99/month per the official US App Store IAP list; OpenAI's own help article (via search-index excerpt) says '$20/month, billed monthly'. OpenAI's web pages block automated fetching, so confirm the live web price in a browser before paying.
Q: Is there an annual ChatGPT Plus plan? A: No — the Plus help article is excerpt-quoted stating that annual billing or multi-month prepay is not supported. That means no prepay discount exists, unlike some $20 competitors.
Q: What usage limits apply on Plus? A: 5-hour rolling windows plus weekly allowances, with limits 'varying based on system conditions' per OpenAI's own wording. The only recent public number found in retrievable form is an excerpted '3,000 messages/week with GPT-5 Thinking' (May 2026). There is no consolidated published Plus cap table.
Q: What happens when I hit a limit? A: Wait for the window to roll, buy flexible-usage credits (personal plans article 12642688), or — for weekly Work/Codex limits — buy an instant reset from Usage settings. Credit and reset prices appear in-product; they are not published on help pages we could retrieve.
Q: Why is this page more hedged than our other offer pages? A: Because every OpenAI-owned URL returned 403 to our fetcher this pass. Where a fact rests on a search-index excerpt rather than a direct read, the Evidence section and each sentence flag it — that is the site's rule: excerpts labeled, never disguised as live verification.
Q: How current is this information? A: App Store listing read 2026-09-07; help-article excerpts surfaced 2026-09-07 with content dating back to May 2026. ChatGPT caps change frequently — re-verify in a browser before paying.
Evidence: [high] Plus priced at $19.99 in the official US App Store In-App Purchases list for OpenAI's ChatGPT app — read directly (server-rendered, Apple-side) 2026-09-07; the same list shows Go $8.00 and Pro $100/$200. (Source: apps.apple.com ChatGPT listing, accessed 2026-09-07) | [medium] Monthly-only billing, auto-renew at standard rate, and 'broader model and tool access than Free' come from quotes the search index attributes to official help articles (6950777, 12803158, 9275245) — the pages themselves returned 403 to direct fetch; browser re-verification required. (Source: Search-index excerpts of help.openai.com, surfaced 2026-09-07) | [low] The 3,000-messages/week GPT-5-Thinking figure and 196k context are attributed to the release-notes article but date from May 2026 indexing — OpenAI moves caps monthly, so treat as historical, not current. (Source: help.openai.com/6825453 release-note excerpt, surfaced 2026-09-07) | [medium] Credit top-ups (article 12642688), instant weekly-limit resets (20001507), and localized multi-currency billing (10421635) are excerpt-attributed to official articles; all dollar-level specifics (credit prices, refund windows) were not retrievable. (Source: help.openai.com article excerpts, surfaced 2026-09-07)
URL: https://token-perks.com/best/chatgpt-plus/
## Google AI Pro subscription [active]
Provider: Google | Plan: Google AI Pro
Price now: $19.99/mo (annual billing is offered but the annual price is not published on official pages fetched)
Renewal: Auto-renews at the start of each billing cycle until you unsubscribe; price increases require at least 30 days' prior notice applied after the current period; a failed renewal gets a seven-day grace period.
Verified: 2026-09-07
Verdict: Google AI Pro is less an AI subscription than a bundle: $19.99/mo buys 5 TB of storage, roughly 4x Gemini usage vs non-subscribers, Gemini across Gmail/Docs, Flow, NotebookLM, Jules, Antigravity, Home Premium, and $10/mo of Cloud credits. For anyone already inside Google Workspace for personal life, it is the best value-per-dollar of the $20 class even if you rarely prompt. For AI-first power users the ceiling is real: a weekly cap over each product, no published numbers, and a credits top-up whose price you discover at checkout. Students get a free year — that alone flips the comparison for eligible users.
Access: Subscribe from gemini.google/subscriptions (or one.google.com) with a personal Google account in a supported country; manage seat, family members, and AI credits from the Google One dashboard.
Caveats: The weekly limit sits above the 5-hour window: bursty days are survivable, sustained weeks are the actual ceiling | Annual billing exists but its price is never shown on official pages fetched this pass — you discover it at checkout | AI-credit packs: mechanism official, pricing unpublished, and per-feature costs are stated to be modifiable by Google | Cancel means you keep AI until cycle end, then storage drops to the 15 GB free tier — if your data lives on the 5 TB, downgrading is a migration project
Limits: Usage vs free tier: '4x access to Gemini' (footnoted 'in comparison to non-AI subscribers'); higher caps for Gemini 3.1 Pro, Deep Research, 1M-token context, ~1,500-page uploads | No published per-day counts for prompts, Deep Research runs, or Flow generations — only the 5-hour-refresh/weekly-limit mechanism | Only the plan manager can purchase AI credits; credits are family-pooled and may expire after a period | Exact credit costs per feature: not published (vary by feature)
Eligibility: Available in over 150 countries and territories; personal Google accounts only (not while billed via a third-party partner or Pixel Pass) | Some benefits are age-gated to 18+ | Students: Google AI Pro free for 1 year with SheerID verification — offer auto-renews at $19.99/mo when student status concludes; not available on school-issued accounts | Family sharing: up to five other members get AI benefits at no extra cost (with carve-outs, e.g. YouTube Premium Lite not shareable)
Restrictions: Limits are per-window: 'Your limit refreshes every 5 hours until you reach your weekly limit' — and each product (Gemini, Flow, Antigravity…) has its own limits | Exact prompts/day or gens/week figures: not published on fetched official pages | Buying more usage exists (AI credits) but pack prices and per-feature credit costs are not published in the help article and 'Google reserves the right to modify the credit cost' | Feature flags are region-limited: auto-browse in Chrome and Photos Remix are US-only per the plans page | The one.google.com plans pages render prices client-side — dollar figures were confirmed only on gemini.google/subscriptions
Two ways to read the $19.99: as an AI plan it must beat Claude Pro/ChatGPT Plus on model access per dollar; as a bundle it competes with a storage plan plus several app subscriptions. The page's own extras (5 TB ≈ standalone paid storage tier, YouTube Premium Lite, Home Premium at '$10/mo value', $10/mo Cloud credits) are what make the sub-$20 sticker defensible.
Against the $0.80/task illustrative reference, $19.99 is parity at ~25 heavy AI tasks/mo — but unlike pure chat subs, part of the fee is storage and apps you may pay for separately anyway; assign that share before comparing AI-only.
Student free year makes marginal cost $0 for a year — the break-even logic only matters from month 13.
The unpublished annual price and credit-pack prices mean true cost comparison against annual-priced rivals (Claude $17/mo annual) cannot be done from official pages this pass.
Q: What does Google AI Pro cost as of the Sep 2026 snapshot? A: $19.99/month — verified in the static text of gemini.google/subscriptions on 2026-09-07. Annual billing is offered but no annual price appears on any official page fetched this pass.
Q: What usage limits does the plan actually state? A: '4x access to Gemini' vs non-subscribers, and the published mechanism: a limit that refreshes every 5 hours until you reach a weekly limit, applied per product. Exact prompt or generation counts are not published.
Q: Can I buy more usage when I hit the limit? A: Yes via Google's AI credits (plain-English: prepaid usage packs for Gemini, Flow, and Antigravity). Only the plan manager can buy them, they pool across the family, they may expire, and per-feature credit costs are unpublished and changeable by Google.
Q: What's in the bundle beyond chat? A: 5 TB storage (10 TB on some variants), Gemini in Gmail/Docs/Vids, Deep Research and 1M-context access, Flow/Flow Music, NotebookLM, Jules, Antigravity, Colab/AI Studio, Android Studio integration, Chrome auto-browse (US), YouTube Premium Lite, Google Home Premium Standard, and $10/mo Cloud credits — all per the official plans/help pages.
Q: How do cancellation and refunds work? A: Cancel anytime; benefits run to the end of the billing cycle, then storage reverts to 15 GB. Purchases are non-refundable in most regions (partial periods included); Apple-billed subscriptions refund through Apple; a seven-day grace period applies to failed renewals.
Q: How current is this information? A: Live fetches of official Google pages 2026-09-07. Region availability and credit pricing shift often — re-verify before paying.
Evidence: [high] The $19.99/mo price, '4x higher usage limits', the 5-hour/weekly limit wording, and 'No refunds for partial billing periods, except as required by law' are in the static text of the official Gemini subscriptions page fetched live. (Source: gemini.google/subscriptions/, accessed 2026-09-07) | [high] Renewal (charge at cycle start, auto-renew), 30-day price-increase notice, seven-day failed-renewal grace, and family/billing mechanics come from Google One's official terms and help articles. (Source: one.google.com/terms-of-service + support.google.com/googleone answers 2736362/9056360/16476811, accessed 2026-09-07) | [high] AI credits as the overage path (Pro/Ultra only, plan-manager purchase, family-pooled, expiring, modifiable costs) is documented officially; pack prices are absent from the doc itself. (Source: support.google.com/googleone/answer/16287445, accessed 2026-09-07) | [medium] 150+ country availability, student 1-year offer terms (SheerID; renews at $19.99; no school accounts), and the 5 TB / 4x / feature lists are from the official AI-plans pages; third-party claims of a ~$49.99 10 TB tier were NOT confirmed on official pages and are not quoted. (Source: one.google.com/about/google-ai-plans/ + intl/en_us variant, accessed 2026-09-07)
URL: https://token-perks.com/best/google-ai-pro/
## Claude Pro subscription [active]
Provider: Anthropic | Plan: Claude Pro
Price now: $20/mo billed monthly; annual is $200 billed up front ('$17 per month with annual subscription discount', pricing-page card)
Renewal: Renews automatically to the payment method on file. Cancel any time and keep access until the current billing period ends — 'to avoid the next charge, cancel at least 24 hours before your renewal date.' Payments are generally non-refundable except where Consumer Terms or local law say otherwise; EEA/UK buyers get a prorated 14-day refund window.
Verified: 2026-09-07
Verdict: Claude Pro is the standard $20 seat for the Claude apps plus Claude Code: Opus-class usage is included (Free is locked out of Opus), the 200k-context models share a 5-hour session + weekly cap system, and everything beyond the caps has a published valve — usage credits at standard API rates with a $2,000/day redemption ceiling and an optional monthly spend cap. Two details decide fit: if you want Fable-class models, Pro meters them per-token from the start; and if you live in Claude Code, terminal and chat drain one pool, so plan weekly capacity, not daily. Annual prepay ($200) trims the effective rate to $17/mo and credits a mid-year switch — the best deal for steady users who can commit 12 months.
Access: Subscribe at claude.com/pricing (web billing, local currency where supported); Claude Code then runs on the same account and pool. When a cap bites, enable usage credits in settings (web) at standard API rates or wait for the 5-hour/weekly reset.
Caveats: The flagship Fable models cost extra-usage credits on Pro — the flat fee covers Opus/Sonnet/Haiku-class limits, not the top of the ladder | One pool for everything: a long Claude Code agent session eats the chat allowance for the same 5-hour window (and the same weekly cap) | Weekly caps reset on an account-assigned schedule, not the calendar — plan around Settings > Usage, not Mondays | Credits are real money at API rates with a $2,000/day redemption limit — heavy metered weeks can dwarf the $20 base | App-store sign-ups inherit store billing: refunds at Apple's discretion, and requesting one does not cancel the subscription
Limits: Session: 'at least 5x' Free usage per rolling 5-hour window (exact token thresholds per model: not published) | Weekly: an all-model weekly limit layered on top; discretion caps 'such as weekly and monthly caps and model and feature usage' possible | Included models: Opus (Pro 'Yes' vs Free 'No'), Sonnet and Haiku on all plans; 200k context window per the matrix | Usage credits: standard API rates, billed separately, $2,000/day redemption limit, optional auto-reload; Japan: credits expire six months after purchase starting September 10, 2026 | Pro does not include Claude Console API usage; an ANTHROPIC_API_KEY in the environment routes Claude Code to API billing instead of the plan | App Store vs web price deltas: not published
Eligibility: Pro is available in Anthropic's supported locations (the supported-countries list spans roughly 190 countries; some regions excluded) and 'pricing in your local currency where supported' | Web sign-up bills directly; iOS/Android sign-ups are billed through Apple/Google | Team products require 2+ seats; Pro does not include Claude Console (API) usage | Mobile-app subscribers can only enable usage credits on the web version of Claude
Restrictions: Caps are window-shaped: 'Pro gives you at least 5x more usage per 5-hour session than Free', the session limit 'resets every five hours', and paid plans add a weekly all-model cap that resets at a fixed time assigned to your account (visible in Settings > Usage) | No fixed message count: usage depends on conversation length/complexity, model, and features; Anthropic 'may limit your usage in other ways… at our discretion' | Fable 5/5.1 (flagship tier) are not inside Pro's included limits — on Pro they 'run on pay-as-you-go usage credits' at standard API rates; Max seats include Fable up to 50% of weekly limits | Claude Code shares one pool with chat: 'All Claude Code activity counts toward your weekly limits — web, desktop, mobile, and terminal draw from the same pool' | Model training on consumer plans (Free/Pro/Max) is 'Opt-out' per the pricing matrix — check the setting rather than assuming a default | 'Price and plans are subject to change at Anthropic's discretion.'
Against the site's illustrative $0.80/task reference, $20/mo is parity at ~25 heavy tasks/month and annual ($200) at ~21 — below that, Free's throttled usage or raw API pay-as-you-go costs less; above it, the flat fee wins as long as the weekly cap never binds.
Annual switch math is stated officially: billed $200 up front, unused monthly time credited to the first annual invoice — mid-year upgrades do not double-pay.
The credits valve converts overage into API-rate spend: if your metered weeks average over ~$80 of credits at API rates, the article's own ladder (Max 5x at $100/mo, itemized in the Max-plan help article) is the cheaper seat — the pricing page's advice is to move up 'if you find yourself hitting usage limits'.
Fable-on-credits means the flagship model is never flat-rate below Max: a steady Fable workload is an API bill wearing a subscription.
Q: What does Claude Pro cost as of the Sep 2026 snapshot? A: $20/month billed monthly, or $200/year billed upfront — the pricing-page card states '$17 Per month with annual subscription discount ($200 billed up front).' Verified live on claude.com/pricing 2026-09-07; displayed prices vary by region and may include tax.
Q: What are the usage limits on Pro? A: 'At least 5x more usage per 5-hour session than Free,' with the session limit resetting every five hours, plus an all-model weekly cap that resets at a fixed, account-assigned time. No fixed message count is published — the official position is that usage depends on conversation size, model, and features.
Q: Is Claude Code included? A: Yes — 'Claude Code is included in all paid plans,' and it shares one pool with chat: terminal and web activity both draw on the same 5-hour session and weekly limits. A separate ANTHROPIC_API_KEY routes Claude Code to Console API billing instead, outside the subscription.
Q: Can I use the Fable models on Pro? A: Only on metered credits: official docs state that on Pro plans, Fable 5/5.1 'run on pay-as-you-go usage credits' at standard API rates (Max includes Fable up to 50% of weekly limits). The free-Fable promo ended July 19, 2026.
Q: What happens when I hit a limit? A: Officially: wait for the reset, upgrade to Max, or enable usage credits (billed at standard API rates, separate on your bill, capped at $2,000/day redemption with an optional monthly spend cap and auto-reload). Note mobile-app subscribers can only enable credits on the web.
Q: How do cancellation and refunds work? A: Cancel anytime; access runs to the period end; cancel ≥24 hours before renewal to avoid the next charge. Payments are generally non-refundable except under Consumer Terms or local law — EEA/UK buyers get a prorated 14-day window; App Store refunds are Apple's call and don't cancel the subscription. Annual early-exit terms beyond upfront billing: not published.
Q: How current is this information? A: Live reads of claude.com/pricing plus nine official help-center articles on 2026-09-07 (some pages 403 the automated fetcher and were re-read via browser-mode request the same day). Caps and pricing are explicitly 'subject to change' — re-verify before paying.
Evidence: [high] Prices ($20/mo; $200/yr '$17 per month' card), the Opus/Sonnet/Haiku/Fable matrix rows, the 5-hour/weekly cap FAQ wording, 'cancel 24 hours before renewal', annual upfront billing, and the 'Model training: Opt-out' row were read on claude.com/pricing and re-verified against saved raw HTML the same day. (Source: claude.com/pricing, accessed 2026-09-07) | [high] Weekly-limit mechanics (account-assigned reset, Settings > Usage), the ≥5x framing, discretion caps, local-currency/multi-region pricing, 'no Console API on Pro', and 'subject to change at Anthropic's discretion' come from official support articles (8325606, 11647753, 11049762). (Source: support.claude.com articles, accessed 2026-09-07) | [high] Usage-credit terms — standard API rates, separate bill line, $2,000/day redemption limit, monthly cap + auto-reload, mobile-enable-on-web-only, Japan 6-month expiry — are quoted from the manage-usage-credits article; Fable-on-credits for Pro and the July 19, 2026 free-Fable promo end from article 15424964. (Source: support.claude.com/en/articles/12429409 + /15424964, accessed 2026-09-07) | [medium] Exact token thresholds behind session/weekly caps, annual early-exit terms, and App Store vs web price deltas: checked across pricing page and eight help articles — not published. (Source: Absence check across claude.com/pricing + support.claude.com, 2026-09-07)
URL: https://token-perks.com/best/claude-pro/
## Cursor Pro editor subscription [active]
Provider: Cursor | Plan: Cursor Pro
Price now: $20/mo (annual USD price not published; a Monthly/Yearly toggle exists in-app but yearly amounts are not on official pages fetched)
Renewal: 'Your cycle starts on the date you subscribe. It renews monthly or yearly based on your plan.' (billing doc, quoted). Canceling keeps paid features to the end of the current period with no further charges.
Verified: 2026-09-07
Verdict: Cursor Pro is the $20 entry to the editor most agents use: unlimited tab completions, a real Agent allowance, and a hard-but-published escalation path (on-demand at API rates, or Pro+ at 3x limits for $60). The unforgiving part is opacity — the included pool's dollar size is not published, so 'included usage' can feel larger than it is on frontier models, and the refund rule (14 days AND unused) means a single heavy afternoon can make the month non-refundable. For completion-heavy coding it is excellent value; for long autonomous agent sessions, read your monthly usage dashboard, not the marketing tab.
Access: Install Cursor, subscribe Pro from cursor.com/pricing (or in-app), decide explicitly whether to enable on-demand usage in dashboard billing settings, and check cursor.com/dashboard for pool consumption.
Caveats: Included-usage pool size: not published — only 'a set amount' — so break-even math vs raw API use requires your own dashboard numbers | On-demand is opt-in but silent about totals until the invoice: billed in arrears at API rates | Refunds require the period to be essentially unused — trying the plan hard for one day forfeits the 14-day refund right | Prices exclude tax; India users get a different SKU (Start ₹649) — regional pricing is thin (one tier), not broad
Limits: Hobby (free): 'Limited Agent requests', no card, Access to Composer — exact request counts not published | Pro: unlimited tab completions; Agent pool + bonus pool included (sizes unpublished) | Pro+ = '3x Pro limits on Agent'; Ultra = '20x Pro limits on Agent' + priority features (help table $60/$200) | On-demand enabled: usage continues at per-model API list prices, monthly in arrears | Throughput/rate caps beyond the pools: not published
Eligibility: Any individual with a supported card ('all major credit and debit cards'); invoice/wire billing exists for larger accounts | All prices exclude applicable taxes; tax details are managed in the Stripe billing portal | India-only 'Start' tier (₹649/mo tax-inclusive) is the regional alternative to Pro | The legacy free-Pro student offer closed to new sign-ups on June 25 2026 — existing redeemers keep their rate until expiry
Restrictions: Included usage: 'Every plan includes a set amount of model usage' — the dollar size of Pro's pool is not published anywhere on official pages fetched | On-demand usage (plain-English: pay-as-you-go at raw model prices once the pool is gone) must be enabled and is billed in arrears 'at the same API rates' | Auto mode bills 'at the list price of the model each request is routed to'; a named third-party model draws the Other Models pool at its API rate — so 'Auto' is not a cheap default for long agent runs | Two separate pools exist (Agent and bonus/other); Pro/Pro+/Ultra include both | Upgrading credits only the unused included usage of the old plan, not time left on it
Because the included pool's dollar size is unpublished, the only honest break-even is empirical: run your normal month with on-demand off, read included-usage consumption, and only then judge whether $20 covered it.
Reference point: at the site's illustrative $0.80/task PAYG figure, $20 covers ~25 heavy tasks — but Cursor bills on-demand at true API list prices, so frontier-model tasks above the pool cost what the model costs, not what the plan implies.
Pro+ ($60, 3x Agent limits) vs three months of Pro + modest on-demand: if on-demand charges run over ~$40/mo consistently, Pro+ is the ladder step the docs themselves point to ('upgrade to a higher plan').
Ultra at $200 buys 20x Pro Agent limits — a per-task price that only makes sense above ~150+ heavy agent tasks/mo against the $0.80 reference; below that the pool ceiling rarely binds on Pro.
Q: What does Cursor Pro cost as of the Sep 2026 snapshot? A: $20/month, verified in live static text on both the pricing page and the help-center table on 2026-09-07. Yearly billing exists in-app but no yearly dollar amount is published on official pages.
Q: How much usage is actually included? A: Cursor states 'a set amount of model usage' across two pools (Agent and bonus/other) but does not publish the dollar size — check your dashboard after a typical week. Beyond the pool, on-demand usage bills at raw API list rates in arrears if enabled.
Q: Can I get a refund? A: Only within 14 days of the charge AND if you have not used the subscription during the period; consumed on-demand/fast-request charges are never refundable. In practice, trial the plan lightly or not at all if you want the refund to remain possible.
Q: Is the student offer still free Pro? A: No — the legacy free-Pro student discount stopped new sign-ups on June 25, 2026. Students who already redeemed keep their rate until it expires; Cursor now points to credits/discounts at campus events instead.
Q: What is the difference between Pro, Pro+, and Ultra? A: Official framing is multiples of Pro's Agent limits: Pro+ = 3x, Ultra = 20x (plus priority access to new features). Tab completions are unlimited on all three. Teams tiers ($40/$120) add per-user admin pricing with Premium at 5x Standard.
Q: How current is this information? A: All quotes from eight official Cursor pages fetched live 2026-09-07. Note the pricing page's static-render behavior has varied between passes — re-verify the number before paying.
Evidence: [high] Pro $20/mo, the ladder (Hobby Free, Pro+ $60, Ultra $200, Teams $40/$120, Start ₹649 India), 'on-demand usage… billed in arrears', unlimited tab completions on Pro/Ultra, and Auto-mode list-price billing were read live on cursor.com pricing and docs pages. (Source: cursor.com/pricing + /help/account-and-billing/pricing + /docs/models-and-pricing + /help/models-and-usage/usage-limits, accessed 2026-09-07) | [high] Renewal-from-subscribe-date, cancel-keeps-access, and the 14-day-and-unused refund rule with non-refundable consumed usage are verbatim from the billing/cancel/refunds docs. (Source: cursor.com/help/account-and-billing/{billing,cancel,refunds}, accessed 2026-09-07) | [high] The student free-Pro discontinuation (June 25 2026 for new sign-ups; existing redeemers honored until expiry) is stated on the official student-discount page. (Source: cursor.com/help/account-and-billing/student-discount, accessed 2026-09-07) | [low] Dollar size of Pro's included usage pool: not published on official pages; third-party figures ($20/$70/$400 pools) exist but are unverified and are NOT quoted as fact. (Source: Absence across cursor.com docs, 2026-09-07)
URL: https://token-perks.com/best/cursor-pro/
## Perplexity Pro subscription [active]
Provider: Perplexity | Plan: Perplexity Pro
Price now: $20/mo; annual billing exists but its price is not published on any official page fetched (the 'about $200/yr' figure rests on an expired India promo's dollar-value line, not a listed annual price)
Renewal: 'Unless you cancel before the trial ends, your subscription will auto-renew at the standard rate for the plan you selected (monthly or annual).' Canceling ends service at the billing-cycle close; no pause option exists.
Verified: 2026-09-07
Verdict: Perplexity Pro is the answer-engine subscription: every frontier chat model in one tab plus search-native citations, at $20/mo with no hard published message counter — limits are 'average use' bands that tighten only in extreme weeks. The trade-offs are opacity (no numeric caps, a broken canonical model list) and thin agent economics (no standing monthly credit pool on consumer Pro). If your workflow is research-and-cite, it is the most natural $20; if it is long agentic runs, the credit math quietly pushes you toward Max or a metered API.
Access: Subscribe from the Pro page or in the Perplexity app (web checkout shows the annual toggle price in-account); then track credit burn under Settings → Billing if you use agentic features.
Caveats: 'Weekly limits (average use)' is the official wording — the cap exists but its number does not, and heavy weeks can throttle | The annual price is invisible on official pages: you see it only at checkout, and the $200/yr figure in third-party listings is promo-valuation lineage | One-time 4,000 bonus credits expire after 30 days — an onboarding grant, not an allowance | Refund windows are short and regional (24 h standard, 72 h annual) — and one official page disagrees with another on the annual window (72 vs 48 hours)
Limits: Pro Search: numeric Pro cap not published (Free = 3/day) | Research queries: Pro = monthly 'average use' limit, number not published; Max = 50/month | Files: up to 50 uploads per project; PDFs/CSV/audio/video/images; higher upload rate limits | Media: 'enhanced image generation', 'limited access to video generation' — counts not published | Computer credits: no set monthly amount on consumer Pro; Max 10,000/mo
Eligibility: Any individual with a supported payment method (App Store and Google Play also sell Pro) | Students: Education Pro at $10/mo with SheerID verification (half price, per the which-plan article) | Max at $200/mo or $2,000/yr — annual Max is web-only; Enterprise Pro from $40/seat/mo, edu/NPO $30/seat | Old promo routes are closed: the Airtel 12-month offer ended January 16 2026
Restrictions: Pro Search: Free gets 3/day; Pro gets 'weekly limits (average use)' — no numeric cap published, and access 'may be limited during weeks of especially heavy usage' | Deep Research on Pro: 'monthly limits (average use)', no figure (Max's 50/month is stated by contrast) | Credits (plain-English: prepaid units for agentic 'Computer' tasks, 100 credits = $1): consumer Pro does not start with a fixed monthly pool — Pro shows only a one-time 4,000 bonus credit grant that expires after 30 days; Max carries 10,000/mo | The official model-list help article that the Pro page links is 404 — 'which models, exactly, today' has no single current canonical page | Agentic features (Labs, Comet agent work) may consume credits beyond the flat fee
Perplexity prices breadth (all frontier models + premium data sources PitchBook/Statista/S&P per the hub FAQ) rather than volume — against the $0.80/task illustrative reference, $20 is parity at ~25 heavy tasks, but no published cap tells you whether 30 tasks is allowed or throttled.
The student $10/mo Education Pro halves the entry price and is the only officially priced discount path we found for individuals.
Agent usage is the hidden variable: 100 credits = $1, Pro has no standing pool, so a week of agentic work is billed by credit purchase, not by the plan — Max's $200 with 10,000/mo credits is the pooled alternative (~$100 of credit face value monthly).
Annual prepay cannot be evaluated on price-per-month because the annual figure is not published — treat any '$200/yr' claim as unverified until checkout shows it.
Q: What does Perplexity Pro cost as of the Sep 2026 snapshot? A: $20/month, read on the official pricing hub 2026-09-07. Annual billing exists but its price is not published on official pages — read the real figure at checkout before prepaying.
Q: What are Pro's usage limits? A: Officially numeric limits are not published: Pro Search carries 'weekly limits (average use)' (Free gets 3/day), Deep Research 'monthly limits (average use)' (Max gets 50/month). Throttling can tighten 'during weeks of especially heavy usage'.
Q: Do Pro credits roll over or reset? A: Consumer Pro has no standing monthly credit pool — the 4,000-credit grant is one-time and expires after 30 days. Credits (100 = $1) are prepaid units for agentic tasks; Max is the tier with a monthly 10,000-credit allowance.
Q: Can I get a refund? A: Regionally: EU/UK/Turkey within 14 days, South Korea/Brazil within 7, elsewhere 24 hours for monthly and 72 hours for annual purchases (one official page says 48 for annual — the pages conflict). Canceled plans keep access until the cycle ends; there is no pause option.
Q: Is there a cheaper official route? A: Education Pro at $10/mo with SheerID verification. The Airtel India 12-month promo (valued at $200) ended January 16 2026 and is not a current path.
Q: How current is this information? A: Pricing hub and five help-center articles fetched 2026-09-07 (help pages self-date July 16–Sep 2 2026). The site 403-blocks standard crawlers, so re-verification needs a browser pass — which is also why the annual figure stays 'not published' here.
Evidence: [high] $20/mo Pro, the Free/Pro/Max feature matrix, 4,000 one-time bonus credits, and premium data sources (PitchBook, Statista, S&P Capital IQ) were read from the official pricing hub's server-rendered text via browser-UA fetch (the automated fetcher is 403-blocked site-wide). (Source: perplexity.ai/hub/pricing, accessed 2026-09-07) | [high] Limits wording ('Pro Search: weekly limits (average use)', 'Research Queries: monthly limits (average use)', Free '3/day', 50 files/project), the model examples (GPT-5.2, Claude Sonnet 4.6, Gemini 3.1 Pro, Grok 4, Sonar), and auto-renew/no-pause mechanics come from five official help-center articles dated Jul 16–Sep 2 2026. (Source: perplexity.ai/help-center articles 10352901/11187416/10353002/10354288/13838041, accessed 2026-09-07) | [medium] Annual Pro price: the billing FAQ proves annual billing exists ('switching from annual to monthly'), but no fetched official page states its USD amount; the search snippet '$17/month billed annually' was not visible on the JS-shell /pro page and is treated as unconfirmed. (Source: Billing FAQ + absence check across hub/pricing, 2026-09-07) | [medium] Refund windows (14d EU/UK/TR, 7d KR/BR, 24h monthly, 72h annual) are from the Refunds article; the Max article states a 48-hour annual window — an unresolved internal conflict between two official pages both dated Jul 16 2026. (Source: help-center 10354288 vs 11680686, accessed 2026-09-07)
URL: https://token-perks.com/best/perplexity-pro/
## Full price table (129 ways to buy tokens, snapshot 2026-09-07)
Every tracked access route follows. Blended $/M = (3 x input + 1 x output) / 4 where both rates are published. Evidence labels: DIRECT = read on the provider's own page; EXCERPT = official copy via snapshot or search index; UNCERTAIN = not verified this pass.
- Anthropic | Subscriptions | Claude Free | List: $0/mo | DIRECT | Source: https://claude.com/pricing | Caveats: Throttled usage | No Opus or Fable models | Page: https://token-perks.com/providers/anthropic/
- Anthropic | Subscriptions | Claude Pro | List: $20/mo ($17/mo annual) | DIRECT | Source: https://claude.com/pricing | Caveats: 5-hour session caps | Fable requires usage credits | Page: https://token-perks.com/providers/anthropic/
- Anthropic | Subscriptions | Claude Max 5x | List: $100/mo | DIRECT | Source: https://support.claude.com/en/articles/11049741-what-is-the-max-plan | Caveats: Web price; mobile may differ | Weekly cap on Fable | Page: https://token-perks.com/providers/anthropic/
- Anthropic | Subscriptions | Claude Max 20x | List: $200/mo | DIRECT | Source: https://support.claude.com/en/articles/11049741-what-is-the-max-plan | Page: https://token-perks.com/providers/anthropic/
- Anthropic | Subscriptions | Claude Team - Standard seat | List: $20/seat/mo annual ($25 monthly) | DIRECT | Source: https://claude.com/pricing | Caveats: Annual billing for the $20 rate | Page: https://token-perks.com/providers/anthropic/
- Anthropic | Subscriptions | Claude Team - Premium seat | List: $100/seat/mo annual ($125 monthly) | DIRECT | Source: https://claude.com/pricing | Page: https://token-perks.com/providers/anthropic/
- Anthropic | Subscriptions | Claude Enterprise | List: $20/seat + usage at API rates | DIRECT | Source: https://claude.com/pricing | Caveats: Usage billed at API rates | Page: https://token-perks.com/providers/anthropic/
- OpenAI | Subscriptions | ChatGPT Plus | List: $19.99/mo | DIRECT | Source: https://apps.apple.com/us/app/chatgpt/id6448311069 | Caveats: App Store price; web may differ | No annual plan (help excerpt) | Caps 'vary based on system conditions' | Page: https://token-perks.com/providers/openai/
- OpenAI | Subscriptions | ChatGPT Go | List: $8.00/mo | DIRECT | Source: https://apps.apple.com/us/app/chatgpt/id6448311069 | Page: https://token-perks.com/providers/openai/
- OpenAI | Subscriptions | ChatGPT Pro 5x | List: $100/mo | DIRECT | Source: https://apps.apple.com/us/app/chatgpt/id6448311069 | Page: https://token-perks.com/providers/openai/
- OpenAI | Subscriptions | ChatGPT Pro 20x | List: $200/mo | DIRECT | Source: https://apps.apple.com/us/app/chatgpt/id6448311069 | Page: https://token-perks.com/providers/openai/
- OpenAI | Subscriptions | ChatGPT Business seat | List: not fetched | UNCERTAIN | Source: https://web.archive.org/web/20260729/https://help.openai.com/en/articles/11487671-flexible-pricing-for-the-enterprise-edu-and-business-plans | Caveats: Per-seat price not published | Page: https://token-perks.com/providers/openai/
- Google | Subscriptions | Google AI Plus | List: $4.99/mo | EXCERPT | Source: https://one.google.com/intl/en_us/about/google-ai-plans/ | Page: https://token-perks.com/providers/google/
- Google | Subscriptions | Google AI Pro | List: $19.99/mo | DIRECT | Source: https://gemini.google/subscriptions/ | Caveats: Annual price not published | Limits: 5-hour refresh to a weekly ceiling, per product | Family sharing up to 5 others | Page: https://token-perks.com/providers/google/
- Google | Subscriptions | Google AI Ultra | List: $99.99/mo | EXCERPT | Source: https://one.google.com/intl/en_us/about/google-ai-plans/ | Page: https://token-perks.com/providers/google/
- xAI | Subscriptions | Grok Free | List: $0/mo | DIRECT | Source: https://x.ai/grok | Caveats: Rate limits unpublished | Page: https://token-perks.com/providers/xai/
- xAI | Subscriptions | SuperGrok | List: $30/mo | EXCERPT | Source: https://x.ai/pricing | Caveats: Official page blocked; search-snippet copy | Page: https://token-perks.com/providers/xai/
- xAI | Subscriptions | SuperGrok Plus | List: $100/mo | EXCERPT | Source: https://x.ai/pricing | Caveats: Official page blocked; search-snippet copy | Page: https://token-perks.com/providers/xai/
- xAI | Subscriptions | SuperGrok Lite | List: $10/mo | DIRECT | Source: https://apps.apple.com/us/app/grok-ai/id6670324846 | Page: https://token-perks.com/providers/xai/
- xAI | Subscriptions | SuperGrok Heavy | List: $300/mo | DIRECT | Source: https://apps.apple.com/us/app/grok-ai/id6670324846 | Caveats: Annual-effective figure unconfirmed | Page: https://token-perks.com/providers/xai/
- Mistral | Subscriptions | Vibe Free | List: $0 + $10/mo API credits | EXCERPT | Source: https://mistral.ai/pricing | Page: https://token-perks.com/providers/mistral/
- Mistral | Subscriptions | Vibe Pro | List: $14.99/mo | EXCERPT | Source: https://mistral.ai/pricing | Page: https://token-perks.com/providers/mistral/
- Mistral | Subscriptions | Vibe Team | List: $24.99 per user/mo | EXCERPT | Source: https://mistral.ai/pricing | Page: https://token-perks.com/providers/mistral/
- Mistral | Subscriptions | Vibe Enterprise | List: Custom | EXCERPT | Source: https://mistral.ai/pricing | Caveats: Custom pricing | Page: https://token-perks.com/providers/mistral/
- Moonshot AI | Subscriptions | Kimi K3 Membership - Moderato | List: $19/mo (annual ~ $15 eff.) | EXCERPT | Source: https://www.kimi.ai/membership/pricing | Caveats: Shared credit pool | 5-hour and weekly controls | Page: https://token-perks.com/providers/moonshot/
- Moonshot AI | Subscriptions | Kimi K3 Membership - Allegretto | List: $39/mo (annual ~ $31 eff.) | EXCERPT | Source: https://www.kimi.ai/membership/pricing | Caveats: Shared credit pool | 5-hour and weekly controls | Page: https://token-perks.com/providers/moonshot/
- Moonshot AI | Subscriptions | Kimi K3 Membership - Allegro | List: $99/mo (annual ~ $79 eff.) | EXCERPT | Source: https://www.kimi.ai/membership/pricing | Caveats: Shared credit pool | 5-hour and weekly controls | Page: https://token-perks.com/providers/moonshot/
- Moonshot AI | Subscriptions | Kimi K3 Membership - Vivace | List: $199/mo (annual ~ $159 eff.) | EXCERPT | Source: https://www.kimi.ai/membership/pricing | Caveats: Shared credit pool | 5-hour and weekly controls | Page: https://token-perks.com/providers/moonshot/
- Z.ai | Subscriptions | GLM Coding Plan - Lite | List: $18/mo | DIRECT | Source: https://docs.z.ai/devpack/overview | Caveats: Credit-cycle caps | Page: https://token-perks.com/providers/zai/
- Z.ai | Subscriptions | GLM Coding Plan - Pro | List: not fetched | UNCERTAIN | Source: https://docs.z.ai/devpack/overview | Caveats: Price not published | Page: https://token-perks.com/providers/zai/
- Z.ai | Subscriptions | GLM Coding Plan - Max | List: not fetched | UNCERTAIN | Source: https://docs.z.ai/devpack/overview | Caveats: Price not published | Page: https://token-perks.com/providers/zai/
- MiniMax | Subscriptions | Token Plan - Plus | List: $22/mo (promo ~ $20) | EXCERPT | Source: https://platform.minimax.io/docs/guides/pricing-token-plan | Caveats: Weekly quota caps | Page: https://token-perks.com/providers/minimax/
- MiniMax | Subscriptions | Token Plan - Max | List: $55/mo (promo ~ $50) | EXCERPT | Source: https://platform.minimax.io/docs/guides/pricing-token-plan | Caveats: Weekly quota caps | Page: https://token-perks.com/providers/minimax/
- MiniMax | Subscriptions | Token Plan - Ultra | List: $132/mo (promo ~ $120) | EXCERPT | Source: https://platform.minimax.io/docs/guides/pricing-token-plan | Caveats: Weekly quota caps | Page: https://token-perks.com/providers/minimax/
- DeepSeek | Subscriptions | No consumer subscription | List: n/a | DIRECT | Source: https://api-docs.deepseek.com/quick_start/pricing | Page: https://token-perks.com/providers/deepseek/
- Perplexity | Subscriptions | Perplexity Pro | List: $20/mo (annual not published) | DIRECT | Source: https://www.perplexity.ai/pro | Caveats: Annual price not published | Caps are 'weekly/monthly limits (average use)' — numbers not published | Page: https://token-perks.com/providers/perplexity/
- Perplexity | Subscriptions | Perplexity Max | List: $200/mo | EXCERPT | Source: https://www.perplexity.ai/hub/pricing | Caveats: Credit caps | Page: https://token-perks.com/providers/perplexity/
- Perplexity | Subscriptions | Perplexity Education Pro | List: $10/mo | EXCERPT | Source: https://www.perplexity.ai/hub/pricing | Caveats: Student verification | Page: https://token-perks.com/providers/perplexity/
- Meta | Subscriptions | No first-party consumer subscription | List: n/a | UNCERTAIN | Source: https://www.together.ai/pricing | Caveats: First-party route unverified | Page: https://token-perks.com/providers/meta/
- Meta | API (per token) | Llama 4 (hosted/first-party) | List: not fetched | UNCERTAIN | Source: https://developer.meta.com/ai/ | Caveats: Not verifiable this pass | Page: https://token-perks.com/providers/meta/
- Anthropic | API (per token) | Fable 5.1 | List: $10 in / $50 out | Blended $20.00/M | DIRECT | Source: https://claude.com/pricing | Caveats: High output rate | Page: https://token-perks.com/providers/anthropic/
- Anthropic | API (per token) | Opus 5 | List: $5 in / $25 out | Blended $10.00/M | DIRECT | Source: https://claude.com/pricing | Page: https://token-perks.com/providers/anthropic/
- Anthropic | API (per token) | Sonnet 5 | List: $2 in / $10 out | Blended $4.00/M | DIRECT | Source: https://claude.com/pricing | Page: https://token-perks.com/providers/anthropic/
- Anthropic | API (per token) | Haiku 4.5 | List: $1 in / $5 out | Blended $2.00/M | DIRECT | Source: https://claude.com/pricing | Page: https://token-perks.com/providers/anthropic/
- Anthropic | API (per token) | Legacy page ladder (Fable 5; Opus 4.x; Sonnet 4.x) | List: $10/$50; $5/$25; $3/$15; $15/$75 as listed | DIRECT | Source: https://claude.com/pricing | Caveats: Superseded ladder; per-model rows not split | Page: https://token-perks.com/providers/anthropic/
- OpenAI | API (per token) | gpt-5.6-sol | List: $4.00 in / $20.00 out | Blended $8.00/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Promo expiry 2026-11-21 | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-5.6-terra | List: $2.00 in / $12.00 out | Blended $4.50/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-5.6-luna | List: $0.20 in / $1.20 out | Blended $0.45/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-6-astra (newest flagship) | List: $10.00 in / $50.00 out | Blended $20.00/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Newest flagship; premium rate | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-5.5 / 5.4 / 5.4-mini / 5.4-nano | List: $5/$30; $2.50/$15; $0.75/$4.50; $0.20/$1.25 | Blended $11.25/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Ladder row; models not split | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-5.2 / gpt-5.1 | List: $1.75/$14; $1.25/$10 | Blended $4.81/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Two models on one row | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-5-mini / gpt-5-nano | List: $0.25/$2; $0.05/$0.40 | Blended $0.69/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Two models on one row | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-5-pro / 5.2-pro / 5.5-pro | List: $15/$120; $21/$168; $30/$180 | Blended $41.25/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Three models on one row; no cache rate | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | o-series (o3 / o3-pro / o4-mini / o1 / o1-pro) | List: $2/$8; $20/$80; $1.10/$4.40; $15/$60; $150/$600 | Blended $3.50/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Five models on one row | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | gpt-4.1 / 4.1-mini / 4.1-nano | List: $2/$8; $0.40/$1.60; $0.10/$0.40 | Blended $3.50/M | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Three models on one row | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | Embeddings (3-small / 3-large / ada-002) | List: $0.02; $0.13; $0.10 | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Input-only; not chat models | Page: https://token-perks.com/providers/openai/
- OpenAI | API (per token) | Batch / Fast mode / residency modifiers | List: Batch -50%; Fast mode ~2x; +10% residency uplift | DIRECT | Source: https://developers.openai.com/api/docs/pricing | Caveats: Modifiers, not standalone access | Page: https://token-perks.com/providers/openai/
- Google | API (per token) | Gemini 3.8 / 3.7 / 3.6 Flash | List: $0.75 in / $3.75 out (thru Dec 31 2026); $1.50 / $7.50 from Jan 1 2027 | Blended $1.50/M | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Caveats: Scheduled price increase 2027-01-01 | Page: https://token-perks.com/providers/google/
- Google | API (per token) | Gemini 3.5 Flash / 3.5 Flash-Lite | List: $1.50/$9.00; $0.30/$2.50 | Blended $3.38/M | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Caveats: Two models on one row | Page: https://token-perks.com/providers/google/
- Google | API (per token) | Gemini 3.1 Flash-Lite | List: $0.25 in / $1.50 out | Blended $0.56/M | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Page: https://token-perks.com/providers/google/
- Google | API (per token) | Gemini 3.1 Pro Preview | List: $2.00 in / $12.00 out (200k and under) | Blended $4.50/M | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Caveats: Long-context surcharge above 200k | No free tier | Page: https://token-perks.com/providers/google/
- Google | API (per token) | Gemini 3 Flash Preview | List: $0.50 in / $3.00 out | Blended $1.13/M | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Caveats: Preview tier | Page: https://token-perks.com/providers/google/
- Google | API (per token) | Gemini 2.5 Pro | List: $1.25 in / $10 out (200k and under) | Blended $3.44/M | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Caveats: Long-context surcharge above 200k | Page: https://token-perks.com/providers/google/
- Google | API (per token) | Gemini 2.5 Flash / 2.5 Flash-Lite | List: $0.30/$2.50; $0.10/$0.40 | Blended $0.85/M | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Caveats: Two models on one row | Page: https://token-perks.com/providers/google/
- xAI | API (per token) | grok-4.6 | List: $2.00 in / $6.00 out (under 200k) | Blended $3.00/M | DIRECT | Source: https://docs.x.ai/docs/models | Caveats: Whole-request repricing above 200k | Page: https://token-perks.com/providers/xai/
- xAI | API (per token) | grok-4.5 | List: $2.00 in / $6.00 out (under 200k) | Blended $3.00/M | DIRECT | Source: https://docs.x.ai/docs/models | Caveats: Whole-request repricing above 200k | Page: https://token-perks.com/providers/xai/
- xAI | API (per token) | grok-4.3 | List: $1.25 in / $2.50 out (under 200k) | Blended $1.56/M | DIRECT | Source: https://docs.x.ai/docs/models | Caveats: Whole-request repricing above 200k | Page: https://token-perks.com/providers/xai/
- xAI | API (per token) | grok-4.20-0309 (reasoning / non-reasoning / multi-agent) | List: Same tiers as grok-4.3 | Blended $1.56/M | DIRECT | Source: https://docs.x.ai/docs/models | Caveats: Logprobs unsupported | Page: https://token-perks.com/providers/xai/
- xAI | API (per token) | grok-build-0.1 (coding) | List: $1.00 in / $2.00 out (under 200k) | Blended $1.25/M | DIRECT | Source: https://docs.x.ai/docs/models | Caveats: Whole-request repricing above 200k | Page: https://token-perks.com/providers/xai/
- xAI | API (per token) | Imagine images / video / voice | List: $0.02-0.05 per image; $0.05-0.08 per sec; $0.08 per min | DIRECT | Source: https://docs.x.ai/docs/models | Caveats: Non-token billing | Page: https://token-perks.com/providers/xai/
- DeepSeek | API (per token) | deepseek-v4-flash | List: Cache-miss in $0.22 off-peak / $0.44 peak; out $0.66 / $1.32 | Blended $0.33/M | DIRECT | Source: https://api-docs.deepseek.com/quick_start/pricing/ | Caveats: Peak/off-peak split - see hours | Page: https://token-perks.com/providers/deepseek/
- DeepSeek | API (per token) | deepseek-v4-pro | List: Cache-miss in $0.66 off-peak / $1.32 peak; out $1.98 / $3.96 | Blended $0.99/M | DIRECT | Source: https://api-docs.deepseek.com/quick_start/pricing/ | Caveats: Peak/off-peak split - see hours | Page: https://token-perks.com/providers/deepseek/
- Moonshot AI | API (per token) | kimi-k3 (API) | List: $3.00 in (cache-miss) / $15.00 out | Blended $6.00/M | DIRECT | Source: https://platform.kimi.ai/docs/pricing/chat-k3 | Caveats: Taxes excluded | Page: https://token-perks.com/providers/moonshot/
- Moonshot AI | API (per token) | Kimi K2.7 Code / K2.7-highspeed / K2.6 (API) | List: K2.7 Code: cache-miss in $0.95 / out $4.00; highspeed $1.90 / $8.00; K2.6 $0.95 / $4.00 | Blended $1.71/M | DIRECT | Source: https://platform.kimi.ai/docs/pricing/chat-k27-code | Caveats: Cache-hit rates differ - see notes | Page: https://token-perks.com/providers/moonshot/
- Z.ai | API (per token) | GLM-5.3 / GLM-5.2 / GLM-5.1 | List: $1.40 in / $4.40 out | Blended $2.15/M | DIRECT | Source: https://docs.z.ai/guides/overview/pricing | Caveats: Context windows unpublished | Page: https://token-perks.com/providers/zai/
- Z.ai | API (per token) | GLM-5.3-Flash | List: $0.075 in / $0.25 out (promo; list $0.15 / $0.50) | Blended $0.12/M | DIRECT | Source: https://docs.z.ai/guides/overview/pricing | Caveats: Blend priced at promo $0.075/$0.25 — list-price blend $0.2375/M | Promo ends 2026-09-09 24:00 UTC+8 | Page: https://token-perks.com/providers/zai/
- Z.ai | API (per token) | GLM-5 / 4.7 / 4.6 / 4.5 ladder | List: $1/$3.2; $0.6/$2.2 (4.7/4.6/4.5) | Blended $1.55/M | DIRECT | Source: https://docs.z.ai/guides/overview/pricing | Caveats: Ladder row; models not split | Page: https://token-perks.com/providers/zai/
- Z.ai | API (per token) | GLM-4.7-FlashX / Flash / 4.5-Flash | List: $0.07/$0.4; Free; Free | Blended $0.15/M | DIRECT | Source: https://docs.z.ai/guides/overview/pricing | Caveats: Zero-priced tiers | Page: https://token-perks.com/providers/zai/
- Z.ai | API (per token) | GLM-4.6V / 4.5V / GLM-OCR / 4.6V-Flash | List: $0.3/$0.9; $0.6/$1.8; $0.03/$0.03; Free | Blended $0.45/M | DIRECT | Source: https://docs.z.ai/guides/overview/pricing | Caveats: Ladder row; models not split | Page: https://token-perks.com/providers/zai/
- Alibaba (Qwen) | API (per token) | Qwen official API (Model Studio) | List: qwen3.8-max $2.00/$6.00; qwen3.7-max $2.50/$7.50; qwen3.7-plus $0.40/$1.60; qwen3.8-flash $0.15/$0.47 | Blended $3.00/M | DIRECT | Source: https://www.alibabacloud.com/help/en/model-studio/model-pricing | Caveats: Region/term tiers - see notes | Page: https://token-perks.com/providers/alibaba/
- Alibaba (Qwen) via Together AI | API (per token) | Qwen3.8-2.4T-A95B / Qwen3.8 Flash / Qwen3.7-Max / Qwen3.7-Plus | List: $2.00/$6.00; $0.15/$0.47; $1.25/$3.75; $0.32/$1.28 | Blended $3.00/M | DIRECT | Source: https://www.together.ai/pricing | Caveats: Resale pricing, not first-party | Page: https://token-perks.com/providers/together/
- Alibaba (Qwen) via Together AI | API (per token) | Qwen3.6-Plus / Qwen3.5-397B-A17B / Qwen3.5-9B / Qwen3-235B-A22B | List: $0.50/$3.00; $0.60/$3.60; $0.17/$0.25; $0.20/$0.60 | Blended $1.13/M | DIRECT | Source: https://www.together.ai/pricing | Caveats: Resale pricing, not first-party | Page: https://token-perks.com/providers/together/
- Meta (Llama) via Together AI | API (per token) | Llama 3.3 70B / Llama 3 8B Instruct Lite | List: $1.04/$1.04; $0.14/$0.14 | Blended $1.04/M | DIRECT | Source: https://www.together.ai/pricing | Caveats: Llama 4 pricing unverified | Page: https://token-perks.com/providers/together/
- Mistral | API (per token) | Mistral Large (API) | List: $0.5 in / $1.5 out | Blended $0.75/M | EXCERPT | Source: https://mistral.ai/pricing | Caveats: Full table not rendered | Page: https://token-perks.com/providers/mistral/
- MiniMax | API (per token) | MiniMax-M2.7 | List: $0.30 in / $1.20 out | Blended $0.52/M | EXCERPT | Source: https://platform.minimax.io/docs/guides/pricing-paygo | Page: https://token-perks.com/providers/minimax/
- MiniMax | API (per token) | MiniMax-M2.7-highspeed | List: $0.60 in / $2.40 out | Blended $1.05/M | EXCERPT | Source: https://platform.minimax.io/docs/guides/pricing-paygo | Caveats: Priority multipliers raise price | Page: https://token-perks.com/providers/minimax/
- Groq | API (per token) | Groq API (console list) | List: gpt-oss-120b $0.15/$0.60; 20b $0.075/$0.30; qwen3.6-27b $0.60/$3.00; qwen3.8-27b $0.80/$4.00 | DIRECT | Source: https://console.groq.com/docs/models | Caveats: Multi-model row - no single blend | Flagship rows sales-gated | Page: https://token-perks.com/providers/groq/
- OpenAI | Credits / prepaid | ChatGPT credit packs | List: 100 credits $4.00; 500 credits $20.00; 1,000 credits $40.00 | DIRECT | Source: https://apps.apple.com/us/app/chatgpt/id6448311069 | Page: https://token-perks.com/providers/openai/
- OpenRouter | Credits / prepaid | Prepaid USD credits | List: 5.5% top-up fee (Stripe, $0.80 min); 5% via crypto (USDC) | EXCERPT | Source: https://openrouter.ai/docs/faq | Caveats: Top-up fee | Credit expiry after 1 year | Page: https://token-perks.com/providers/openrouter/
- OpenRouter | Credits / prepaid | Free models tier | List: $0 | EXCERPT | Source: https://openrouter.ai/docs/faq | Caveats: Daily request caps | Page: https://token-perks.com/providers/openrouter/
- NVIDIA Build | Credits / prepaid | Kimi K3 free (dev/prototyping) | List: $0 within account limits | EXCERPT | Source: https://build.nvidia.com | Caveats: Account-variable limits | Dev/prototyping scope only | Page: https://token-perks.com/providers/nvidia-build/
- Fireworks AI | Credits / prepaid | Serverless PAYG + $1 starter credit | List: Postpaid per token; $1 free credits for new accounts | DIRECT | Source: https://fireworks.ai/pricing | Caveats: 10 RPM cap without payment method | Page: https://token-perks.com/providers/fireworks/
- Fireworks AI | Credits / prepaid | Fire Pass | List: No listed price; promo-code activated; invite only | DIRECT | Source: https://docs.fireworks.ai/firepass | Caveats: Invite-only | Non-production only | Expiration only visible in-account | Page: https://token-perks.com/providers/fireworks/
- Fireworks AI | Credits / prepaid | Training / GPU | List: LoRA SFT from $0.50 per 1M training tokens (16B and under) to $10 (300B+); H100 $7-8/GPU-hr; B200 $10-13 | DIRECT | Source: https://fireworks.ai/pricing | Caveats: Region surcharge 1.5x | Page: https://token-perks.com/providers/fireworks/
- Together AI | Credits / prepaid | Serverless PAYG (no credit packs) | List: Per-token postpaid; H100 dedicated $3.99/hr; B200 $8.19/hr | DIRECT | Source: https://www.together.ai/pricing | Page: https://token-perks.com/providers/together/
- Cerebras | Credits / prepaid | Free credits + Developer tier | List: $5 free credits; Developer self-serve from $10 | EXCERPT | Source: https://www.cerebras.ai/pricing | Caveats: Per-model prices not rendered | Page: https://token-perks.com/providers/cerebras/
- DeepInfra / Hyperbolic / Novita / Lambda | Credits / prepaid | Third-party hosting (observed, not rowed) | List: DeepInfra seen at $1.74/$3.48 for deepseek-v4-pro in search results | Blended $2.17/M | UNCERTAIN | Source: https://deepinfra.com/ | Caveats: Unverified third-party figure | Page: https://token-perks.com/providers/deepinfra/
- Cursor | Coding tools | Hobby | List: Free | DIRECT | Source: https://cursor.com/pricing | Page: https://token-perks.com/providers/cursor/
- Cursor | Coding tools | Pro | List: $20/mo | DIRECT | Source: https://cursor.com/pricing | Caveats: On-demand usage billed in arrears | Included pool size not published | Page: https://token-perks.com/providers/cursor/
- Cursor | Coding tools | Cursor Pro+ | List: $60/mo | DIRECT | Source: https://cursor.com/help/account-and-billing/pricing | Page: https://token-perks.com/providers/cursor/
- Cursor | Coding tools | Cursor Ultra | List: $200/mo | DIRECT | Source: https://cursor.com/help/account-and-billing/pricing | Page: https://token-perks.com/providers/cursor/
- Cursor | Coding tools | Teams Standard / Premium | List: $40/user/mo (Standard); $120/user/mo (Premium) | DIRECT | Source: https://cursor.com/help/account-and-billing/pricing | Page: https://token-perks.com/providers/cursor/
- GitHub | Coding tools | Copilot Free | List: $0 | DIRECT | Source: https://github.com/features/copilot/plans | Caveats: Monthly request caps | Page: https://token-perks.com/providers/github/
- GitHub | Coding tools | Copilot Pro | List: $10/user/mo | DIRECT | Source: https://github.com/features/copilot/plans | Caveats: Credit-capped premium models | Page: https://token-perks.com/providers/github/
- GitHub | Coding tools | Copilot Pro+ | List: $39/user/mo | DIRECT | Source: https://github.com/features/copilot/plans | Caveats: Credit-capped premium models | Page: https://token-perks.com/providers/github/
- GitHub | Coding tools | Copilot Max | List: $100/user/mo | DIRECT | Source: https://github.com/features/copilot/plans | Caveats: Credit-capped premium models | Page: https://token-perks.com/providers/github/
- GitHub | Coding tools | Copilot Business | List: $19/seat/mo | DIRECT | Source: https://docs.github.com/en/copilot/get-started/plans | Page: https://token-perks.com/providers/github/
- GitHub | Coding tools | Copilot Enterprise | List: $39/seat/mo | DIRECT | Source: https://docs.github.com/en/copilot/get-started/plans | Page: https://token-perks.com/providers/github/
- GitHub | Coding tools | AI Credits unit (cross-cutting) | List: 1 AI credit = $0.01 | DIRECT | Source: https://github.com/features/copilot/plans | Caveats: Unit definition, not a plan | Page: https://token-perks.com/providers/github/
- Windsurf / Devin (Cognition) | Coding tools | Devin pricing (post-merge ladder) | List: Free $0; Pro $20/mo; Max $200/mo; Teams $80 + $40/seat/mo; Enterprise custom | EXCERPT | Source: https://web.archive.org/web/20260814065709/https://devin.ai/pricing | Caveats: Via official-page snapshot - re-verify live | Page: https://token-perks.com/providers/devin/
- Google | Coding tools | Antigravity (agentic IDE/CLI/SDK) | List: Available at no charge | DIRECT | Source: https://antigravity.google | Caveats: Limits unpublished | Page: https://token-perks.com/providers/antigravity/
- Google | Coding tools | Jules (agentic coding) | List: $0 free tier; paid tiers bundle into Google AI Pro/Ultra | DIRECT | Source: https://jules.google/docs/usage-limits/ | Caveats: No standalone price - bundled | Page: https://token-perks.com/providers/google/
- OpenAI | Coding tools | Codex via ChatGPT plans / API | List: Bundled in ChatGPT plans (no standalone price) | EXCERPT | Source: https://web.archive.org/web/20260906/https://help.openai.com/en/articles/11481834-chatgpt-rate-card-business-enterpriseedu-credit-based-pricing | Caveats: Availability rolling out - re-verify | Page: https://token-perks.com/providers/openai/
- Moonshot AI | Coding tools | Kimi coding endpoint (membership route) | List: Membership tiers $19-199 (category a) | EXCERPT | Source: https://www.kimi.ai/membership/pricing | Caveats: Shared credit pool | Page: https://token-perks.com/providers/moonshot/
- Z.ai | Coding tools | GLM Coding Plan (tool coverage) | List: $18+ (see category a) | DIRECT | Source: https://docs.z.ai/devpack/overview | Caveats: Credit-cycle caps | Page: https://token-perks.com/providers/zai/
- Google | Free / promos | Gemini API free tier (AI Studio) | List: $0 | DIRECT | Source: https://ai.google.dev/gemini-api/docs/pricing | Caveats: Request-rate caps | Page: https://token-perks.com/providers/google/
- NVIDIA Build | Free / promos | Kimi K3 dev/prototyping | List: $0 | EXCERPT | Source: https://build.nvidia.com | Caveats: Account-variable limits | Page: https://token-perks.com/providers/nvidia-build/
- Meta Muse Spark via OpenCode Zen | Free / promos | Muse Spark 1.3 Contributor Free | List: $0 for input, cache, and output | EXCERPT | Source: https://opencode.ai/docs/zen/ | Caveats: In-product activation | End date unpublished | Page: https://token-perks.com/providers/muse-spark/
- Mistral | Free / promos | Vibe Free | List: $0 + $10/mo API credits | EXCERPT | Source: https://mistral.ai/pricing | Page: https://token-perks.com/providers/mistral/
- Mistral | Free / promos | Student Pro pricing | List: $5.99/mo (normally $14.99) | EXCERPT | Source: https://mistral.ai/pricing | Caveats: Student verification | Page: https://token-perks.com/providers/mistral/
- OpenRouter | Free / promos | New-user allowance + free models | List: $0 | EXCERPT | Source: https://openrouter.ai/docs/faq | Caveats: Daily request caps | Page: https://token-perks.com/providers/openrouter/
- Cerebras | Free / promos | Free trial | List: $5 free credits | EXCERPT | Source: https://www.cerebras.ai/pricing | Caveats: One-time credit | Page: https://token-perks.com/providers/cerebras/
- Fireworks AI | Free / promos | Starter credit | List: $1 free credits | DIRECT | Source: https://fireworks.ai/pricing | Caveats: One-time credit | Page: https://token-perks.com/providers/fireworks/
- Fireworks AI | Free / promos | Fire Pass | List: $0 per-token on included models (invite-only pass) | DIRECT | Source: https://docs.fireworks.ai/firepass | Caveats: Invite-only | Non-production only | Page: https://token-perks.com/providers/fireworks/
- xAI | Free / promos | Grok free tier | List: $0 | DIRECT | Source: https://x.ai/grok | Caveats: Rate limits unpublished | Page: https://token-perks.com/providers/xai/
- Anthropic | Free / promos | Claude Free | List: $0 | DIRECT | Source: https://claude.com/pricing | Caveats: Throttled usage | Page: https://token-perks.com/providers/anthropic/
- Z.ai | Free / promos | Free API models | List: $0 | DIRECT | Source: https://docs.z.ai/guides/overview/pricing | Caveats: Zero-priced models | Page: https://token-perks.com/providers/zai/
- Z.ai | Free / promos | Night-owl promo (coding plans) | List: $0 within plan window | DIRECT | Source: https://docs.z.ai/devpack/overview | Caveats: Paid-plan requirement | Time-windowed | Page: https://token-perks.com/providers/zai/
- Google | Free / promos | Antigravity developer access | List: $0 (available at no charge) | DIRECT | Source: https://antigravity.google | Caveats: Limits unpublished | Page: https://token-perks.com/providers/antigravity/
## Providers
Alibaba (Qwen): 1 routes. https://token-perks.com/providers/alibaba/
Anthropic: 13 routes. https://token-perks.com/providers/anthropic/
Cerebras: 2 routes. https://token-perks.com/providers/cerebras/
Cursor: 5 routes. https://token-perks.com/providers/cursor/
DeepSeek: 3 routes. https://token-perks.com/providers/deepseek/
Fireworks AI: 5 routes. https://token-perks.com/providers/fireworks/
GitHub: 7 routes. https://token-perks.com/providers/github/
Google: 12 routes. https://token-perks.com/providers/google/
Google Antigravity: 2 routes. https://token-perks.com/providers/antigravity/
Groq: 1 routes. https://token-perks.com/providers/groq/
Meta (Llama): 2 routes. https://token-perks.com/providers/meta/
MiniMax: 5 routes. https://token-perks.com/providers/minimax/
Mistral: 7 routes. https://token-perks.com/providers/mistral/
Moonshot AI: 7 routes. https://token-perks.com/providers/moonshot/
Muse Spark (via Zen): 1 routes. https://token-perks.com/providers/muse-spark/
NVIDIA Build: 2 routes. https://token-perks.com/providers/nvidia-build/
OpenAI: 19 routes. https://token-perks.com/providers/openai/
OpenRouter: 3 routes. https://token-perks.com/providers/openrouter/
Perplexity: 3 routes. https://token-perks.com/providers/perplexity/
Third-party hosting (observed): 1 routes. https://token-perks.com/providers/deepinfra/
Together AI: 4 routes. https://token-perks.com/providers/together/
Windsurf / Devin (Cognition): 1 routes. https://token-perks.com/providers/devin/
xAI: 12 routes. https://token-perks.com/providers/xai/
Z.ai: 11 routes. https://token-perks.com/providers/zai/
## Guides
Buying AI access as a gift: none of the tracked offers sells a gift card or transferable plan in verified terms — activation, refund windows, and quota reality per offer. /guides/buying-ai-access-as-a-gift/
Effective cost per task, explained: price ÷ tasks done; illustrative $40 basket / 120-task / $0.80 example breaks even at 50 tasks (the calculator itself defaults to the actual $39 Allegretto tier, ≈49).
Monthly vs annual AI plans: annual saves ~20% (Allegretto $39 vs ~$31 effective) for stable volume only.
Cost calculators (/cost-calculator/): break-even with URL-shareable inputs (?tasks=&tokens=&seats=&sub=), one-task cost per priced API route, monthly-vs-annual prepay math, and a 12-month view of every tracked offer.
V0 contains no affiliate links. Re-verify at official terms before paying.