API rate limits and pricing
This page is about the developer APIs, where you pay per token with an API key. It doesn't cover the chat subscriptions in our other guides. Every number below comes from the vendor's own docs, checked 2026-10-16. Prices are in USD per 1 million tokens ("1M tokens") at standard (non-batch) rates.
Rate limits vs. pricing, in plain words
- Pricing is what you pay: a price per 1M input tokens (what you send) and per 1M output tokens (what the model writes). A token is roughly a short word or part of a word. Many vendors charge less for cached input, meaning repeated text the system has already seen.
- Rate limits cap how fast you can send requests, however much you're willing to pay. RPM is requests per minute, TPM is tokens per minute (some vendors split it into input and output: ITPM / OTPM), and RPD is requests per day. Go over a limit and you get an HTTP 429 error; wait and retry.
- Most vendors put you in a tier that rises automatically as you spend more, and higher tiers get higher limits. Some also set a monthly spend cap for each tier.
| Vendor | How limits are measured | How you move up | Free API use? |
|---|---|---|---|
| xAI | Requests per second (RPS) + TPM, per model | Cumulative spend since Jan 1, 2026 (Tier 0–4) | Not stated on the pages we checked |
| Anthropic | RPM + ITPM + OTPM, per model | Automatic, based on usage history (Start, Build, Scale) | "A small amount of free credits" for new users |
| OpenAI | RPM + TPM (also RPD/TPD on some models) | Total credit purchases (Build, Launch, Grow) | Free tier: $100/month usage limit, allowed regions only |
| Google Gemini API | RPM + input TPM + RPD, per project | Billing and spend (Free, Tier 1–3) | Yes, Free tier on many models |
| Meta Model API | RPM + TPM, per team, per model | No tiers published | Not stated on the pages we checked |
xAI (Grok API)
Pricing (docs.x.ai/docs/models, checked 2026-10-16):
| Model | Input | Cached input | Output |
|---|---|---|---|
| grok-4.7 (prompt under 200k tokens) | $2.00 | $0.50 | $6.00 |
| grok-4.7 (prompt 200k tokens or more) | $4.00 | $1.00 | $12.00 |
| grok-4.3 (prompt under 200k tokens) | $1.25 | $0.20 | $2.50 |
| grok-build-0.1 (prompt under 200k tokens) | $1.00 | $0.20 | $2.00 |
Long prompts: once a request reaches the 200k threshold, all tokens in that request are billed at the higher rate.
Rate limits (docs.x.ai rate limits, checked 2026-10-16):
- Your tier depends on cumulative API spend since January 1, 2026: Tier 0 at $0, Tier 1 at $50, Tier 2 at $250, Tier 3 at $1,000, Tier 4 at $5,000, and Enterprise on request. Tiers never downgrade.
- grok-4.7: 150 RPS and 50M TPM at Tier 0, up to 500 RPS and 100M TPM at Tier 4.
- grok-4.3: 37 RPS and 10M TPM at Tier 0, up to 208 RPS and 85M TPM at Tier 4.
- Cached prompt tokens still count toward TPM, even though they're billed at a lower rate.
Anthropic (Claude API)
Pricing (Claude API pricing, checked 2026-10-16):
| Model | Input | Cache hit | Output |
|---|---|---|---|
| Claude Opus 5.5 | $4 | $0.20 | $20 |
| Claude Sonnet 5.5 | $2 | $0.10 | $10 |
| Claude Sonnet 5 | $2 | $0.20 | $10 |
| Claude Haiku 5.5 (prompts up to 100,000 tokens) | $0.10 | $0.01 | $0.50 |
| Claude Haiku 5.5 (prompts over 100,000 tokens) | $0.50 | $0.05 | $2.50 |
| Claude Fable 5.1 | $10 | $0.25 | $50 |
Batch API: 50% off input and output. Claude 4.7 and later models use a newer tokenizer that produces about 30% more tokens for the same text.
Fast mode (up to 2.5x faster output from the same model) costs $8 input / $40 output per 1M tokens on Claude Opus 5.5, and $10 / $50 on Opus 5 and Opus 4.8 (those two are still a research preview). Fast mode can't be combined with the Batch API. On Oct 15, 2026 Anthropic said Opus 5.5 fast mode is out of research preview and no longer has a waitlist (pricing, @ClaudeDevs).
Rate limits (Claude API rate limits, checked 2026-10-16):
| Tier | Monthly spend cap | Opus 5.5 / Sonnet 5.5 / Haiku 5.5 (RPM · ITPM · OTPM) |
|---|---|---|
| Start | $500 | 1,000 · 2,000,000 · 400,000 |
| Build | $1,000 | 5,000 · 5,000,000 · 1,000,000 |
| Scale | $200,000 | 10,000 · 10,000,000 · 2,000,000 |
- For most models, cached input tokens do not count toward ITPM.
- New organizations may start in an Evaluation tier, with lower limits than shown.
- If you hit the monthly spend cap, API use pauses until 00:00 UTC on the first day of the next month, unless you get a higher limit sooner.
OpenAI (API)
Pricing (platform.openai.com/docs/pricing, checked 2026-10-16). Prices are for "short context", meaning up to 272K input tokens:
| Model | Input | Cached input | Output |
|---|---|---|---|
| gpt-6-astra | $10.00 | $1.00 | $50.00 |
| gpt-6.1-sol | $2.00 | $0.10 | $10.00 |
| gpt-6-luna | $0.10 | $0.01 | $0.50 |
| gpt-5.6-terra | $2.00 | $0.20 | $12.00 |
Long-context requests (over 272K input tokens) cost more. Batch and Flex are half price on these models. OpenAI says GPT-5.6 Sol's promotional pricing runs at least through November 21, 2026.
Rate limits (OpenAI rate limits guide, checked 2026-10-16):
| Tier | Qualifies at | Monthly usage limit | Astra, Sol, Terra (RPM · TPM) | Luna (RPM · TPM) |
|---|---|---|---|---|
| Free | Allowed geography | $100 | Not published on this page | Not published on this page |
| Build | $5 total credit purchases | $500 | 5,000 · 1,000,000 | 5,000 · 2,000,000 |
| Launch | $100 | $5,000 | 10,000 · 4,000,000 | 10,000 · 10,000,000 |
| Grow | $500 | $200,000 | 15,000 · 40,000,000 | 30,000 · 180,000,000 |
Limits apply per organization and per project, not per user. Ramping up traffic too fast can trigger a slow_down 429 error even when you're under your limits.
Google (Gemini API)
Pricing (ai.google.dev/gemini-api/docs/pricing, page last updated 2026-10-15 UTC, checked 2026-10-16):
| Model | Free tier | Paid input | Paid output |
|---|---|---|---|
| gemini-4-argon | Not available | $2.00 through Nov 30, 2026; $4.00 from Dec 1, 2026 | $10.00 through Nov 30, 2026; $20.00 from Dec 1, 2026 |
| gemini-3.8-flash | Free of charge | $0.75 through Dec 31, 2026; $1.50 from Jan 1, 2027 | $3.75 through Dec 31, 2026; $7.50 from Jan 1, 2027 |
| gemini-3.5-flash-lite | Free of charge | $0.30 | $2.50 |
| gemini-2.5-flash-lite | Free of charge | $0.10 (text/image/video) | $0.40 |
Output prices include thinking tokens. Google says content sent on the Free tier is used to improve its products, and content sent on the Paid tier isn't.
Rate limits (Gemini API rate limits, checked 2026-10-16):
- Limits are measured in RPM, input TPM and RPD, per project rather than per API key. RPD resets at midnight Pacific time.
- The docs page doesn't publish per-model RPM/TPM numbers. Google says to check your active limits in Google AI Studio.
- Tiers:
- Free: needs an active project.
- Tier 1: needs a linked billing account; $250 billing cap.
- Tier 2: $100 paid plus 3 days; $2,000 cap.
- Tier 3: $1,000 paid plus 30 days; $20,000–$100,000+ cap.
- Spend-based limit per rolling 10 minutes: Tier 1 $10, Tier 2 $50, Tier 3 $200.
Meta (Meta Model API: Muse)
Meta's official API is Meta Model API (base URL api.meta.ai/v1), which serves Muse Wonder and Muse Spark (overview, checked 2026-10-16). The open-weight models (Muse Spark Open and Muse Glimmer) are self-hosted, so the API prices below don't apply to them.
Pricing (Meta pricing and rate limits, checked 2026-10-16):
| Model | Input | Cached input | Output |
|---|---|---|---|
| muse-wonder-1.0 (Standard mode) | $4.00 | $0.20 | $20.00 |
| muse-spark-1.3 / 1.2 / 1.1 (Standard mode) | $1.25 | $0.15 | $4.25 |
| muse-spark-1.3/1.2-contributor (Contributor mode) | $0.10 | $0.002 | $0.20 |
Contributor mode is much cheaper because Meta may use your prompts and outputs to train future models. Standard mode doesn't allow that. Meta says there's no extra charge for long context.
Rate limits (same page) apply per team and per model:
- muse-wonder-1.0 and muse-spark-1.3: 3,000 RPM and 4,000,000 TPM.
- Contributor-mode models: 150 RPM and 3,000,000 TPM.
- No usage tiers are published.
Not covered: cloud resellers (Amazon Bedrock, Google Cloud, Azure), which set their own prices. When a vendor changes a number, we'll log it in the change log.