Rate limits - updated October 17, 2026
AI rate limits and usage limits
Developers Digest tracks the published rate limits and usage limits for 19 AI providers and plans (12 model APIs, 3 cloud platforms and 4 coding tool plans), with 338 entries copied verbatim from official documentation. Each page lists every tier, how to raise your limits and what the docs leave out.
Limits are shown per provider and never compared across providers: each one counts requests, tokens and credits on its own terms.
Model APIs
12 pages5 tiers - 17 tables
Anthropic API
Claude Fable 5.x (Messages API), Start tier: 1,000 RPM and 500,000 ITPM
Verified Limits3 tiers - 14 tables
OpenAI API
Astra, Sol, Terra (Standard rate limits, guide table), Build: 5,000 RPM and 1,000,000 TPM
Verified Limits4 tiers - 4 tables
Gemini API
Gemini API, all models (per project; tier set at billing account level), Tier 1: $250 Billing tier cap (monthly spend cap) and $10 Spend rate limit (per 10 minutes, rolling)
Verified Limits5 tiers - 10 tables
xAI API
grok-4.7, grok-4.6, grok-4.5, Tier 0: 150 RPS and 50M TPM
Verified Limits3 tables
Z.ai API
GLM Coding Team Plan (per seat): GLM-5.3, GLM-5.3-Flash, plus MCP tools, Premium Seat: 35,000 5-Hour Credits and 155,000 Weekly Credits
Verified Limits1 table
Moonshot Kimi API
Batch API (kimi-k2.7-code, kimi-k2.6; kimi-k3 not supported): Input file size: no larger than 100MB (.jsonl, non-empty); Models per batch: only one model per batch is allowed
Verified Limits2 tables
Mistral API
Models configured for Priority Tier (service_tier=auto), Priority Tier (enterprise account setup): 99.5% Uptime SLA
Verified Limits3 tiers - 6 tables
Groq
canopylabs/orpheus-arabic-saudi, Developer plan (base limits, as labelled on the page): 10 RPM and 100 RPD
Verified Limits2 tables
Together AI
Provisioned throughput (GLM-5.3, GLM-5.3 Flash, DeepSeek V4.1 Flash, GLM-5.2, Kimi K3), Provisioned throughput (PTU commitment): PTU price: $0.05 per PTU per minute; Minimum term: one month
Verified Limits4 tiers - 6 tables
Fireworks AI
Serverless models with < 600B total parameters (Small tier): DeepSeek V4.1 Flash, GLM 5.3 Flash, MiniMax M3, NVIDIA Nemotron 3 Ultra (Preview), OpenAI GPT OSS 120B, NVIDIA Nemotron 3.5 Lightning 30B A3B, plus unlisted models under 600B, Tier 3 and above (maximum adaptive ceiling): 216M Total Prompt TPM (input, cached + uncached) and 54M Uncached Prompt TPM
Verified Limits4 tables
OpenRouter
Free model variants (model IDs ending in `:free`), Credits purchased (all time): Less than 10: 20 Requests per minute and 50 Requests per day
Verified Limits5 tiers - 6 tables
Cerebras Inference
gpt-oss-120b, Free Trial: 5 RPM and 30K Uncached TPM
Verified Limits
Cloud platforms
3 pages34 tables
Amazon Bedrock
Anthropic Claude Opus 4.8: 30,000,000 Cross-region TPM [adjustable] and 30,000,000 Global cross-region TPM [adjustable]
Verified Limits5 tiers - 28 tables
Vertex AI
Gemini Pro and Argon models, Standard PayGo Tier 2: $250 - $2,000 Customer spend (30 days) and 1,000,000 Traffic TPM (org-level, prioritized throughput threshold)
Verified Limits68 tables
Azure OpenAI
codex-mini, Tier 1: 1,000 RPM and 1,000,000 TPM (GlobalStandard)
Verified Limits
Coding tool plans
4 pages3 tables
Claude plans and Claude Code
Claude Fable 5 and Claude Fable 5.1, All paid plans: Usage-based Enterprise and Claude API: billed at standard API rates
Verified Limits3 tiers - 2 tables
ChatGPT plans and Codex
GPT-6 Astra (Codex, ChatGPT sign-in), Standard Business: 5-45 Local messages per 5h (estimate)
Verified Limits7 tiers - 10 tables
GitHub Copilot
Copilot Free plan (Auto model selection only), Copilot Free: price: Free; models: Auto model selection only
Verified Limits7 tiers - 4 tables
Cursor
Cursor models (Grok 4.7, Grok 4.6, Grok 4.5, Composer 2.5) on The X Subscription, Plus: Price: $8/mo ($80/yr annual); Included usage: Base
Verified Limits
Method
How we verify
Rate limits move faster than most docs pages admit. These pages record what each provider published on a given day, with the gaps called out instead of guessed at.
- Official docs only
- Every value comes from the provider's own documentation, help center or pricing page. No third-party trackers, forum posts or screenshots.
- Copied verbatim
- Values keep the docs' exact wording and units, so 1.2K stays 1.2K and $500 USD stays $500 USD. Nothing is rounded, converted or ranked.
- Read on the date shown
- Each page carries the day its sources were read and links every table to the page it came from. Limits change often, so check the source before you plan capacity.
Newsletter
Know when limits change
One email per week. New model releases, pricing and limit changes, and what they mean for what you build.