# Token Perks
AI subscription offers, compared on effective cost per task.

Research snapshot Sep 6–7 2026 — re-verify at official terms before paying.
Full price table: 129 ways to buy tokens across 24 providers (subscriptions, API per-token, credits, coding tools, free tiers), snapshot 2026-09-07.
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); uncertainty labels DIRECT / EXCERPT / UNCERTAIN on every row; unverified routes are shown as unverified, never estimated from memory.
Artificial Analysis Intelligence Index values are quoted per datum with source links under AA brief-citation terms; they are absent from this file and all machine feeds by policy.

## Offers
- Kimi K3 Membership (active): $19–$199/mo by tier (Moderato $19, Allegretto $39, Allegro $99, Vivace $199). Renewal: Monthly tiers renew at list price. Annual billing lowers the effective monthly cost (about $15 / $31 / $79 / $159 by tier) but is paid upfront — confirm the refund and cancellation terms on the official page before committing. Caveats: Credits are shared across one pool with 5-hour and weekly controls — heavy sprint days can throttle before the month ends. Details: https://token-perks.com/best/kimi-k3-core/
- Muse Spark 1.3 Free (active): $0 for input, cache, and output tokens during the promo window. Renewal: There is no renewal because there is no paid plan here — it is a $0 limited-time promo on input, cache, and output. Treat anything you build on it as rented time: confirm the promo window and the post-promo price before depending on it. Caveats: $0 in, cache, and out — but it is a limited-time promo. When the window closes, the meter starts. Details: https://token-perks.com/best/muse-spark-zen-free/
- NVIDIA K3 Free (active): $0 for development and prototyping use within account limits. Renewal: Nothing renews because nothing is billed in this snapshot — the free allocation for development and prototyping is set by account-level limits that vary. If your prototype becomes production, expect to move to paid terms published in the official console. Caveats: Free for dev and prototyping with reasoning and tool calls preserved — but limits vary by account, so quota must be measured in your own account. Details: https://token-perks.com/best/nvidia-k3-free/
- Copilot Pro (active): $10/user/mo (monthly billing; annual price for Pro not published). Renewal: Billed monthly per user. Canceling keeps access until the end of the current billing cycle with no proration; upgrades prorate immediately, downgrades take effect next cycle. A personal plan is automatically canceled (with prorated refund) if a company seat overlaps it. Whether the price is locked between renewals: not published. Caveats: Completions are unlimited on paid plans, but chat, agents, and code review draw a $15/mo AI-credit pool — premium models drain it fast, and unused credits are forfeited at month end. Details: https://token-perks.com/best/copilot-pro/
- ChatGPT Plus (active): $19.99/mo in the official US App Store IAP list; the Plus help article is excerpt-quoted at '$20/month, billed monthly'. Renewal: Monthly auto-renew at the standard rate unless canceled; app-store sign-ups renew through Apple/Google and must be canceled there. A consumer price-change notice policy for Plus: not published in anything retrievable this pass. Refund mechanics exist as a dedicated help article but specific windows were not readable. Caveats: OpenAI's own pricing pages block automated reads (403) and render prices client-side, so every figure here rests on the official US App Store listing plus dated excerpts of official help articles — strong, but not a live terms read; verify in a browser before paying. Details: https://token-perks.com/best/chatgpt-plus/
- Google AI Pro (active): $19.99/mo (annual billing is offered but the annual price is not published on official pages fetched). Renewal: Charged at the beginning of each billing cycle; renews automatically unless canceled. Google may change prices with at least 30 days' notice before your next charge. Cancellations keep benefits until the cycle ends; storage purchases are non-refundable in most countries (Apple-billed purchases go through Apple; Israel has a 14-day pro-rata rule). Caveats: The $19.99 fee bundles a lot (5 TB storage, 4x Gemini usage, apps) but the AI limits are window-shaped — a 5-hour refresh that stops at a weekly ceiling — and the annual price is never shown on official pages. Details: https://token-perks.com/best/google-ai-pro/
- Claude Pro (active): $20/mo billed monthly; annual is $200 billed up front ('$17 per month with annual subscription discount', pricing-page card). Renewal: Auto-renews monthly (or yearly on annual, which 'is billed once up front for the year'). Switching monthly to annual credits any unused monthly time against the first annual invoice and resets the next billing date one year out. App Store/Google Play subscriptions are billed and refunded by the store ('Refunds for App Store purchases are granted at Apple's discretion' — and a refund request does not cancel the subscription). Beyond the 12-month upfront structure, annual early-exit terms: not published. Caveats: Usage is capped per rolling 5-hour session and per week — hitting a cap needs usage credits (plain-English: pay-as-you-go top-ups at standard API rates), and the flagship Fable models run on those credits for every plan below Max. Details: https://token-perks.com/best/claude-pro/
- Cursor Pro (active): $20/mo (annual USD price not published; a Monthly/Yearly toggle exists in-app but yearly amounts are not on official pages fetched). Renewal: Auto-renews monthly (or yearly where chosen — the yearly dollar figure is not published) from the subscription date. Refunds: eligible only if the charge is under 14 days old AND the subscription was unused in the period; consumed on-demand/fast-request charges are never refundable. Plan-change credit is computed from unused included usage, not days remaining. Caveats: The $20 includes a usage pool, not unlimited frontier models: after it, on-demand billing at raw API rates runs in arrears — and Cursor does not publish the dollar size of the included pool. Details: https://token-perks.com/best/cursor-pro/
- Perplexity Pro (active): $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: Monthly auto-renew at the standard rate unless canceled; annual subscriptions renew at the annual rate (figure not published). Account → Subscription → Manage Plan → Cancel Subscription; cancellations and downgrades take effect at the end of the billing cycle. Regional refund windows: EU/UK/Turkey 14 days; South Korea/Brazil 7 days; elsewhere monthly up to 24 hours, annual 72 hours (one official Max article says 48 — see evidence). Caveats: Pro answers are capped by unstated weekly/monthly numbers — 'average use' is the published unit — and the annual price never appears on official pages; only a $20/mo figure and an expired promo's '$200 value' line anchor the sticker. Details: https://token-perks.com/best/perplexity-pro/

## Cost leaderboard (blended $/M, cheapest first — cost side only)
- Z.ai GLM-5.3-Flash: $0.12/M blended (DIRECT). https://token-perks.com/providers/zai/
- Z.ai GLM-4.7-FlashX / Flash / 4.5-Flash: $0.15/M blended (DIRECT). https://token-perks.com/providers/zai/
- DeepSeek deepseek-v4-flash: $0.33/M blended (DIRECT). https://token-perks.com/providers/deepseek/
- Z.ai GLM-4.6V / 4.5V / GLM-OCR / 4.6V-Flash: $0.45/M blended (DIRECT). https://token-perks.com/providers/zai/
- OpenAI gpt-5.6-luna: $0.45/M blended (DIRECT). https://token-perks.com/providers/openai/
- MiniMax MiniMax-M2.7: $0.52/M blended (EXCERPT). https://token-perks.com/providers/minimax/
- Google Gemini 3.1 Flash-Lite: $0.56/M blended (DIRECT). https://token-perks.com/providers/google/
- OpenAI gpt-5-mini / gpt-5-nano: $0.69/M blended (DIRECT). https://token-perks.com/providers/openai/
- Mistral Mistral Large (API): $0.75/M blended (EXCERPT). https://token-perks.com/providers/mistral/
- Google Gemini 2.5 Flash / 2.5 Flash-Lite: $0.85/M blended (DIRECT). https://token-perks.com/providers/google/
- DeepSeek deepseek-v4-pro: $0.99/M blended (DIRECT). https://token-perks.com/providers/deepseek/
- Meta (Llama) via Together AI Llama 3.3 70B / Llama 3 8B Instruct Lite: $1.04/M blended (DIRECT). https://token-perks.com/providers/together/
- MiniMax MiniMax-M2.7-highspeed: $1.05/M blended (EXCERPT). https://token-perks.com/providers/minimax/
- Google Gemini 3 Flash Preview: $1.13/M blended (DIRECT). https://token-perks.com/providers/google/
- Alibaba (Qwen) via Together AI Qwen3.6-Plus / Qwen3.5-397B-A17B / Qwen3.5-9B / Qwen3-235B-A22B: $1.13/M blended (DIRECT). https://token-perks.com/providers/together/
- Full ranked table: https://token-perks.com/#leaderboard (interactive; also https://token-perks.com/api/leaderboard.json)
- Full price table (all 129 ways to buy tokens, incl. unverified, labeled): https://token-perks.com/universe/
- Break-even crossover story (flat vs metered): https://token-perks.com/crossover/

## 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: https://token-perks.com/guides/buying-ai-access-as-a-gift/
- Effective cost per task, explained: https://token-perks.com/guides/effective-cost-per-task-explained/
- Monthly vs annual AI plans: https://token-perks.com/guides/monthly-vs-annual-ai/
- AI seats for teams (seat math vs pooled API spend): https://token-perks.com/guides/ai-seats-for-teams/
- Cost calculators (break-even, team seats, one-task cost, prepay, 12-month view): https://token-perks.com/cost-calculator/

## Machine feeds
- Full route data (cost side): https://token-perks.com/api/leaderboard.json
- Offer data: https://token-perks.com/api/offers.json
- Full text: https://token-perks.com/llms-full.txt
- Verification log: https://token-perks.com/changes/
- Methodology: https://token-perks.com/methodology/

Re-verify at official terms before paying. V0 contains no affiliate links.