{ "last_updated": "2026-07-31", "schema_version": "1.0", "notes": "RPM = requests/minute, TPM = tokens/minute, RPD = requests/day, TPD = tokens/day. ITPM/OTPM = input/output tokens per minute (Anthropic-style). Null means not published or not applicable. Limits apply per API key unless noted.", "models": [ { "provider": "OpenAI", "provider_id": "openai", "model_id": "gpt-5.5", "model_name": "GPT-5.5", "docs_url": "https://platform.openai.com/docs/guides/rate-limits", "notes": "GPT-5.5 and GPT-5.4 share identical rate limit profiles across all tiers. No free API tier for GPT-5.x models.", "limits": { "tier-1": { "spend_threshold_usd": 5, "rpm": 500, "tpm": 500000, "rpd": null, "tpd": null, "notes": null }, "tier-2": { "spend_threshold_usd": 50, "rpm": 5000, "tpm": 800000, "rpd": null, "tpd": null, "notes": "7+ days since first successful payment required" }, "tier-3": { "spend_threshold_usd": 100, "rpm": 5000, "tpm": 2000000, "rpd": null, "tpd": null, "notes": null }, "tier-5": { "spend_threshold_usd": 1000, "rpm": 15000, "tpm": 40000000, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "OpenAI", "provider_id": "openai", "model_id": "gpt-5.4-nano", "model_name": "GPT-5.4 Nano", "docs_url": "https://platform.openai.com/docs/guides/rate-limits", "notes": "High-volume classification and routing model. Highest published throughput ceiling of any provider at Tier 5.", "limits": { "tier-1": { "spend_threshold_usd": 5, "rpm": 500, "tpm": 200000, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 100, "rpm": 5000, "tpm": 4000000, "rpd": null, "tpd": null, "notes": null }, "tier-5": { "spend_threshold_usd": 1000, "rpm": 30000, "tpm": 180000000, "rpd": null, "tpd": null, "notes": "Highest published TPM of any provider" } } }, { "provider": "Anthropic", "provider_id": "anthropic", "model_id": "claude-opus-4-8", "model_name": "Claude Opus 4.8", "docs_url": "https://docs.anthropic.com/en/api/rate-limits", "notes": "Opus 4.8, 4.7, and 4.6 share one rate limit pool. Cached input tokens do not count toward ITPM. No free API tier; minimum entry is $5.", "limits": { "tier-1": { "spend_threshold_usd": 5, "rpm": 50, "itpm": 20000, "otpm": 4000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-2": { "spend_threshold_usd": 40, "rpm": 1000, "itpm": 160000, "otpm": 32000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 200, "rpm": 2000, "itpm": 800000, "otpm": 160000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-4": { "spend_threshold_usd": 400, "rpm": 4000, "itpm": 2000000, "otpm": 400000, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Anthropic", "provider_id": "anthropic", "model_id": "claude-sonnet-5", "model_name": "Claude Sonnet 5", "docs_url": "https://docs.anthropic.com/en/api/rate-limits", "notes": "Sonnet 5 and Sonnet 4.6 likely share a rate limit pool (not yet formally documented). Sonnet 5 tokenizer emits ~30% more tokens per request than Sonnet 4.6 for identical prompts. Introductory pricing through August 31, 2026.", "limits": { "tier-1": { "spend_threshold_usd": 5, "rpm": 50, "itpm": 30000, "otpm": 8000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-2": { "spend_threshold_usd": 40, "rpm": 1000, "itpm": 200000, "otpm": 40000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 200, "rpm": 2000, "itpm": 800000, "otpm": 160000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-4": { "spend_threshold_usd": 400, "rpm": 4000, "itpm": 2000000, "otpm": 400000, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Anthropic", "provider_id": "anthropic", "model_id": "claude-haiku-4-5", "model_name": "Claude Haiku 4.5", "docs_url": "https://docs.anthropic.com/en/api/rate-limits", "notes": "Highest throughput Anthropic model. Cached input tokens do not count toward ITPM.", "limits": { "tier-1": { "spend_threshold_usd": 5, "rpm": 50, "itpm": 50000, "otpm": 10000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-2": { "spend_threshold_usd": 40, "rpm": 1000, "itpm": 400000, "otpm": 80000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 200, "rpm": 2000, "itpm": 1000000, "otpm": 200000, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-4": { "spend_threshold_usd": 400, "rpm": 4000, "itpm": 4000000, "otpm": 800000, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Google", "provider_id": "google", "model_id": "gemini-3-1-pro", "model_name": "Gemini 3.1 Pro", "docs_url": "https://ai.google.dev/gemini-api/docs/rate-limits", "notes": "Google removed static rate limit tables from public docs in Q1 2026. Figures shown are last confirmed published numbers. Check AI Studio (aistudio.google.com/rate-limit) for current project limits. Gemini 3.5 Flash is now Google's current flagship but has no published limits.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 5, "tpm": null, "rpd": 100, "tpd": null, "notes": "Very conservative; not suitable as a development baseline" }, "tier-1": { "spend_threshold_usd": 0, "rpm": 150, "tpm": 1000000, "rpd": 1000, "tpd": null, "notes": "Pay-as-you-go (billing enabled)" }, "tier-2": { "spend_threshold_usd": 250, "rpm": 1000, "tpm": 2000000, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Google", "provider_id": "google", "model_id": "gemini-3-1-flash-lite", "model_name": "Gemini 3.1 Flash-Lite", "docs_url": "https://ai.google.dev/gemini-api/docs/rate-limits", "notes": "Highest confirmed published throughput on the free tier (250K TPM). Once billing is enabled on a project the free tier disappears entirely.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 15, "tpm": 250000, "rpd": 1000, "tpd": null, "notes": null }, "tier-1": { "spend_threshold_usd": 0, "rpm": 300, "tpm": 2000000, "rpd": 1500, "tpd": null, "notes": "Pay-as-you-go (billing enabled)" }, "tier-2": { "spend_threshold_usd": 250, "rpm": 2000, "tpm": 4000000, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Mistral", "provider_id": "mistral", "model_id": "mistral-api", "model_name": "Mistral API (all models)", "docs_url": "https://docs.mistral.ai/admin/user-management-finops/tier", "notes": "Mistral uses RPS (requests per second), TPM, and tokens/month - not RPM. Actual limits are only visible after login at admin.mistral.ai/plateforme/limits. No public numerical figures. European data residency (GDPR).", "limits": { "free": { "spend_threshold_usd": 0, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Limited rate limits for evaluation and prototyping only. Upgrade to Scale plan to increase." }, "tier-1": { "spend_threshold_usd": 0, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Scale plan (pay-as-you-go). Actual limits in admin console." }, "tier-2": { "spend_threshold_usd": 20, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Automatic upgrade. Actual limits in admin console." }, "tier-3": { "spend_threshold_usd": 100, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Automatic upgrade. Actual limits in admin console." }, "tier-4": { "spend_threshold_usd": 500, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Automatic upgrade. Contact support for higher limits." } } }, { "provider": "Cohere", "provider_id": "cohere", "model_id": "command-r-plus", "model_name": "Command R+", "docs_url": "https://docs.cohere.com/docs/rate-limits", "notes": "Command A+ (May 2026) is the current flagship but its rate limit figures are not yet published. Command R+ figures shown as reference baseline. Trial keys: 1,000 calls/month across all endpoints.", "limits": { "trial": { "spend_threshold_usd": 0, "rpm": 20, "tpm": null, "rpd": 33, "tpd": null, "notes": "~33 RPD reflects 1,000 calls/month total across all endpoints" }, "production": { "spend_threshold_usd": null, "rpm": 500, "tpm": null, "rpd": null, "tpd": null, "notes": "Production API key required" } } }, { "provider": "Groq", "provider_id": "groq", "model_id": "llama-4-scout", "model_name": "Llama 4 Scout (Groq)", "docs_url": "https://console.groq.com/docs/rate-limits", "notes": "Groq runs on custom LPU hardware. Cached tokens do not count toward TPM/TPD. Limits apply at the organization level.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 30, "tpm": 30000, "rpd": 1000, "tpd": 500000, "notes": "No credit card required" }, "developer": { "spend_threshold_usd": null, "rpm": 1000, "tpm": 300000, "rpd": null, "tpd": null, "notes": "Paid plan" } } }, { "provider": "Groq", "provider_id": "groq", "model_id": "llama-4-maverick", "model_name": "Llama 4 Maverick (Groq)", "docs_url": "https://console.groq.com/docs/rate-limits", "notes": "Maverick free tier is half the default Scout limits. Cached tokens do not count toward TPM/TPD.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 15, "tpm": 3000, "rpd": 500, "tpd": 500000, "notes": "Half the default Groq free tier limits" }, "developer": { "spend_threshold_usd": null, "rpm": 1000, "tpm": 250000, "rpd": null, "tpd": null, "notes": "Paid plan" } } }, { "provider": "Together AI", "provider_id": "together-ai", "model_id": "together-ai-api", "model_name": "Together AI API (all models)", "docs_url": "https://docs.together.ai/docs/rate-limits", "notes": "Dynamic rate limits since January 2026. No fixed tiers or published RPM/TPM. Limits grow with sustained usage and are returned in API response headers (X-RateLimit-* fields).", "limits": { "dynamic": { "spend_threshold_usd": null, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Dynamic limits based on sustained usage. Check API response headers for current limits." } } }, { "provider": "Fireworks AI", "provider_id": "fireworks-ai", "model_id": "fireworks-ai-api", "model_name": "Fireworks AI API (open models)", "docs_url": "https://docs.fireworks.ai/guides/quotas_usage/rate-limits", "notes": "Dynamic ceiling up to 6,000 RPM. Soft limit starts at ~1 RPS and doubles hourly. On-demand GPU deployments remove rate limits entirely. Spending tier caps: $50-$50K/month.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 10, "tpm": null, "rpd": null, "tpd": null, "notes": "No payment method required" }, "paid": { "spend_threshold_usd": null, "rpm": 6000, "tpm": null, "rpd": null, "tpd": null, "notes": "Dynamic ceiling, not guaranteed. Soft limit starts ~1 RPS and doubles hourly." } } }, { "provider": "DeepSeek", "provider_id": "deepseek", "model_id": "deepseek-v4-flash", "model_name": "DeepSeek V4 Flash", "docs_url": "https://api-docs.deepseek.com/", "notes": "DeepSeek does not publish fixed RPM/TPM. Limits are fully dynamic based on server load. No minimum guaranteed. Build fallback routing if DeepSeek availability is critical. V3.2 aliases deprecated July 24, 2026.", "limits": { "paid": { "spend_threshold_usd": null, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Dynamic concurrency limits. Not published. Check API response headers." } } }, { "provider": "DeepSeek", "provider_id": "deepseek", "model_id": "deepseek-v4-pro", "model_name": "DeepSeek V4 Pro", "docs_url": "https://api-docs.deepseek.com/", "notes": "DeepSeek does not publish fixed RPM/TPM. Limits are fully dynamic based on server load.", "limits": { "paid": { "spend_threshold_usd": null, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Dynamic concurrency limits. Not published. Check API response headers." } } }, { "provider": "xAI", "provider_id": "xai", "model_id": "grok-4-3", "model_name": "Grok 4.3", "docs_url": "https://docs.x.ai/docs", "notes": "xAI uses a 5-tier structure ($0/$50/$250/$1K/$5K thresholds) but numerical RPM/TPM are only visible after login in the xAI Console. No public figures.", "limits": { "tier-0": { "spend_threshold_usd": 0, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Free tier. Limits in xAI Console." }, "tier-1": { "spend_threshold_usd": 50, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Limits in xAI Console." }, "tier-2": { "spend_threshold_usd": 250, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Limits in xAI Console." }, "tier-3": { "spend_threshold_usd": 1000, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Limits in xAI Console." }, "tier-4": { "spend_threshold_usd": 5000, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "Limits in xAI Console." } } }, { "provider": "Perplexity", "provider_id": "perplexity", "model_id": "sonar", "model_name": "Sonar", "docs_url": "https://docs.perplexity.ai/docs/admin/rate-limits-usage-tiers", "notes": "Perplexity uses 6 tiers (T0-T5) based on cumulative spend. Leaky bucket algorithm. Token prices exclude a per-request fee of $5-14/1K requests.", "limits": { "tier-0": { "spend_threshold_usd": 0, "rpm": 50, "tpm": null, "rpd": null, "tpd": null, "notes": "New account" }, "tier-1": { "spend_threshold_usd": 50, "rpm": 150, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 500, "rpm": 1000, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Perplexity", "provider_id": "perplexity", "model_id": "sonar-pro", "model_name": "Sonar Pro", "docs_url": "https://docs.perplexity.ai/docs/admin/rate-limits-usage-tiers", "notes": "Same tier structure as Sonar. Deep Research model has much lower limits (5-40 RPM).", "limits": { "tier-0": { "spend_threshold_usd": 0, "rpm": 50, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-1": { "spend_threshold_usd": 50, "rpm": 150, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 500, "rpm": 1000, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Amazon Bedrock", "provider_id": "amazon-bedrock", "model_id": "amazon-bedrock-api", "model_name": "Amazon Bedrock API (all models)", "docs_url": "https://docs.aws.amazon.com/bedrock/latest/userguide/quotas.html", "notes": "Per-model, per-region TPM quotas. Amazon Bedrock no longer enforces RPM limits on the bedrock-runtime endpoint - throttling is token-based only (input + output combined). Default quotas vary by model and region and are visible in the AWS Service Quotas console (select Amazon Bedrock). Provisioned Throughput removes rate limits entirely. Hosts Claude, Llama 4, Mistral, and Amazon Nova models.", "limits": { "default": { "spend_threshold_usd": null, "rpm": null, "tpm": null, "rpd": null, "tpd": null, "notes": "No RPM limit enforced. TPM quotas are per-model, per-region, per-account. Check AWS Service Quotas console under Amazon Bedrock for your account's current limits." } } }, { "provider": "Cerebras", "provider_id": "cerebras", "model_id": "cerebras-gpt-oss-120b", "model_name": "GPT-OSS 120B (Cerebras)", "docs_url": "https://inference-docs.cerebras.ai/support/rate-limits", "notes": "Served on Cerebras wafer-scale silicon. Dual-bucket model: uncached TPM (primary limit) and total TPM (3x uncached). Free trial requires adding a payment method to receive $5 credit; expires in 30 days.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 5, "tpm": 30000, "rpd": null, "tpd": 1000000, "notes": "Free trial. Also 1M tokens/hour (TPH) hard cap." }, "developer": { "spend_threshold_usd": null, "rpm": 1000, "tpm": 1000000, "rpd": null, "tpd": null, "notes": "Pay as You Go (first credit purchase). No hourly or daily token caps." } } }, { "provider": "Cerebras", "provider_id": "cerebras", "model_id": "cerebras-glm-4-7", "model_name": "GLM-4.7 (Cerebras)", "docs_url": "https://inference-docs.cerebras.ai/support/rate-limits", "notes": "ZhipuAI GLM-4.7 served on Cerebras hardware. Same dual-bucket token model as other Cerebras endpoints.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 5, "tpm": 30000, "rpd": null, "tpd": 1000000, "notes": "Free trial. Also 1M tokens/hour (TPH) hard cap." }, "developer": { "spend_threshold_usd": null, "rpm": 500, "tpm": 500000, "rpd": null, "tpd": null, "notes": "Pay as You Go. No hourly or daily token caps." } } }, { "provider": "Cerebras", "provider_id": "cerebras", "model_id": "cerebras-gemma-4-31b", "model_name": "Gemma 4 31B (Cerebras)", "docs_url": "https://inference-docs.cerebras.ai/support/rate-limits", "notes": "Google Gemma 4 31B served on Cerebras hardware. Supports images: 2/request on free tier (4 MB payload), 10/request on developer tier (10 MB payload).", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 5, "tpm": 30000, "rpd": null, "tpd": 1000000, "notes": "Free trial. 2 images per request, 4 MB payload limit." }, "developer": { "spend_threshold_usd": null, "rpm": 300, "tpm": 500000, "rpd": null, "tpd": null, "notes": "Pay as You Go. 10 images per request, 10 MB payload limit." } } }, { "provider": "SambaNova", "provider_id": "sambanova", "model_id": "sambanova-llama-3-3-70b", "model_name": "Llama 3.3 70B (SambaNova)", "docs_url": "https://docs.sambanova.ai/docs/en/models/rate-limits", "notes": "Highest RPM on SambaNova at developer tier. All developer-tier accounts share a 20M tokens/day cap across all models. Served on SambaNova RDU hardware with high sustained throughput.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 20, "tpm": null, "rpd": 20, "tpd": 200000, "notes": "No payment method required." }, "developer": { "spend_threshold_usd": null, "rpm": 240, "tpm": null, "rpd": 48000, "tpd": null, "notes": "Payment method linked. 20M TPD cap shared across all models on the account." } } }, { "provider": "SambaNova", "provider_id": "sambanova", "model_id": "sambanova-deepseek-v3-1", "model_name": "DeepSeek V3.1 (SambaNova)", "docs_url": "https://docs.sambanova.ai/docs/en/models/rate-limits", "notes": "DeepSeek V3.1 served on SambaNova RDU hardware. Not the same as the DeepSeek-hosted API; rate limits and latency differ. 20M TPD cap shared across all models at developer tier.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 20, "tpm": null, "rpd": 20, "tpd": 200000, "notes": "No payment method required." }, "developer": { "spend_threshold_usd": null, "rpm": 60, "tpm": null, "rpd": 12000, "tpd": null, "notes": "Payment method linked. 20M TPD cap shared across all models on the account." } } }, { "provider": "NVIDIA NIM", "provider_id": "nvidia-nim", "model_id": "nvidia-nim-api", "model_name": "NVIDIA NIM API (all models)", "docs_url": "https://build.nvidia.com/explore/discover", "notes": "100+ models on build.nvidia.com including Llama, Mistral, Gemma, Phi, and NVIDIA Nemotron. Free tier via NVIDIA Developer Program (free signup, no credit card). Production use requires NVIDIA AI Enterprise license. RPM varies by model and server load.", "limits": { "free": { "spend_threshold_usd": 0, "rpm": 40, "tpm": null, "rpd": null, "tpd": null, "notes": "~1,000 inference credits included. 40 RPM is the community baseline; actual limit varies by model and traffic. TPM not published." } } }, { "provider": "Perplexity", "provider_id": "perplexity", "model_id": "sonar-reasoning-pro", "model_name": "Sonar Reasoning Pro", "docs_url": "https://docs.perplexity.ai/docs/admin/rate-limits-usage-tiers", "notes": "Search-augmented reasoning model. Same tier thresholds as Sonar and Sonar Pro. Tiers 4 and 5 both cap at 4,000 RPM.", "limits": { "tier-0": { "spend_threshold_usd": 0, "rpm": 50, "tpm": null, "rpd": null, "tpd": null, "notes": "New account" }, "tier-1": { "spend_threshold_usd": 50, "rpm": 150, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 500, "rpm": 1000, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Perplexity", "provider_id": "perplexity", "model_id": "sonar-deep-research", "model_name": "Sonar Deep Research", "docs_url": "https://docs.perplexity.ai/docs/admin/rate-limits-usage-tiers", "notes": "Autonomous multi-step research model with significantly lower rate limits due to heavy compute per request. Async endpoint available at POST /v1/async/sonar. Tiers 4 and 5 cap at 60 and 100 RPM respectively.", "limits": { "tier-0": { "spend_threshold_usd": 0, "rpm": 5, "tpm": null, "rpd": null, "tpd": null, "notes": "New account" }, "tier-1": { "spend_threshold_usd": 50, "rpm": 10, "tpm": null, "rpd": null, "tpd": null, "notes": null }, "tier-3": { "spend_threshold_usd": 500, "rpm": 40, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Azure OpenAI", "provider_id": "azure-openai", "model_id": "azure-gpt-5-1", "model_name": "GPT-5.1", "docs_url": "https://learn.microsoft.com/en-us/azure/ai-services/openai/quotas-limits", "notes": "Azure OpenAI uses subscription-level quota pools (not per-deployment) since May 2026. Global Standard shares one pool across all Azure regions. Data Zone Standard splits into US and EU pools. Tiers auto-upgrade (Free -> Tier 1 -> Tier 6) based on usage and account relationship. Values shown are Tier 1 defaults.", "limits": { "global-standard": { "spend_threshold_usd": null, "rpm": 10000, "tpm": 1000000, "rpd": null, "tpd": null, "notes": "Single pool shared across all Azure regions. Tier 1 default." }, "data-zone-standard": { "spend_threshold_usd": null, "rpm": 3000, "tpm": 300000, "rpd": null, "tpd": null, "notes": "One pool per data zone (US zone, EU zone). Tier 1 default." } } }, { "provider": "Azure OpenAI", "provider_id": "azure-openai", "model_id": "azure-gpt-5-4", "model_name": "GPT-5.4", "docs_url": "https://learn.microsoft.com/en-us/azure/ai-services/openai/quotas-limits", "notes": "Azure OpenAI uses subscription-level quota pools since May 2026. Values shown are Tier 1 defaults.", "limits": { "global-standard": { "spend_threshold_usd": null, "rpm": 10000, "tpm": 1000000, "rpd": null, "tpd": null, "notes": "Single pool shared across all Azure regions. Tier 1 default." }, "data-zone-standard": { "spend_threshold_usd": null, "rpm": 300, "tpm": 300000, "rpd": null, "tpd": null, "notes": "One pool per data zone (US zone, EU zone). Tier 1 default." } } }, { "provider": "Azure OpenAI", "provider_id": "azure-openai", "model_id": "azure-gpt-5-4-nano", "model_name": "GPT-5.4 Nano", "docs_url": "https://learn.microsoft.com/en-us/azure/ai-services/openai/quotas-limits", "notes": "Azure OpenAI uses subscription-level quota pools since May 2026. Values shown are Tier 1 defaults.", "limits": { "global-standard": { "spend_threshold_usd": null, "rpm": 5000, "tpm": 5000000, "rpd": null, "tpd": null, "notes": "Single pool shared across all Azure regions. Tier 1 default." }, "data-zone-standard": { "spend_threshold_usd": null, "rpm": 2000, "tpm": 2000000, "rpd": null, "tpd": null, "notes": "One pool per data zone (US zone, EU zone). Tier 1 default." } } }, { "provider": "Azure OpenAI", "provider_id": "azure-openai", "model_id": "azure-o3", "model_name": "o3", "docs_url": "https://learn.microsoft.com/en-us/azure/ai-services/openai/quotas-limits", "notes": "Azure OpenAI uses subscription-level quota pools since May 2026. Values shown are Tier 1 defaults.", "limits": { "global-standard": { "spend_threshold_usd": null, "rpm": 1000, "tpm": 1000000, "rpd": null, "tpd": null, "notes": "Single pool shared across all Azure regions. Tier 1 default." }, "data-zone-standard": { "spend_threshold_usd": null, "rpm": 300, "tpm": 300000, "rpd": null, "tpd": null, "notes": "One pool per data zone (US zone, EU zone). Tier 1 default." } } }, { "provider": "Azure OpenAI", "provider_id": "azure-openai", "model_id": "azure-o4-mini", "model_name": "o4-mini", "docs_url": "https://learn.microsoft.com/en-us/azure/ai-services/openai/quotas-limits", "notes": "Azure OpenAI uses subscription-level quota pools since May 2026. Values shown are Tier 1 defaults. Note: a 300 RPM burst window applies per 10-second interval.", "limits": { "global-standard": { "spend_threshold_usd": null, "rpm": 1000, "tpm": 1000000, "rpd": null, "tpd": null, "notes": "Single pool shared across all Azure regions. Tier 1 default." }, "data-zone-standard": { "spend_threshold_usd": null, "rpm": 300, "tpm": 300000, "rpd": null, "tpd": null, "notes": "One pool per data zone (US zone, EU zone). Tier 1 default. 300 RPM burst window per 10-second interval." } } }, { "provider": "AI21 Labs", "provider_id": "ai21", "model_id": "jamba-large", "model_name": "Jamba Large", "docs_url": "https://docs.ai21.com/docs/rate-limits", "notes": "Limits apply per API key. RPS (10) is also enforced in addition to RPM. TPM not published. Contact sales@ai21.com for higher limits.", "limits": { "default": { "spend_threshold_usd": null, "rpm": 200, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "AI21 Labs", "provider_id": "ai21", "model_id": "jamba-mini", "model_name": "Jamba Mini", "docs_url": "https://docs.ai21.com/docs/rate-limits", "notes": "Limits apply per API key. RPS (10) is also enforced in addition to RPM. TPM not published. Contact sales@ai21.com for higher limits.", "limits": { "default": { "spend_threshold_usd": null, "rpm": 200, "tpm": null, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Qwen", "provider_id": "qwen", "model_id": "qwen3-7-max", "model_name": "Qwen3.7-Max", "docs_url": "https://www.alibabacloud.com/help/en/model-studio/rate-limit", "notes": "Limits are per Alibaba Cloud root account (aggregated across all API keys). International deployment scope. Batch API calls are exempt from rate limits.", "limits": { "default": { "spend_threshold_usd": null, "rpm": 600, "tpm": 1000000, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Qwen", "provider_id": "qwen", "model_id": "qwen3-5-plus", "model_name": "Qwen3.5-Plus", "docs_url": "https://www.alibabacloud.com/help/en/model-studio/rate-limit", "notes": "Limits are per Alibaba Cloud root account (aggregated across all API keys). International deployment scope. Batch API calls are exempt from rate limits.", "limits": { "default": { "spend_threshold_usd": null, "rpm": 15000, "tpm": 6000000, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Qwen", "provider_id": "qwen", "model_id": "qwen3-5-flash", "model_name": "Qwen3.5-Flash", "docs_url": "https://www.alibabacloud.com/help/en/model-studio/rate-limit", "notes": "Limits are per Alibaba Cloud root account (aggregated across all API keys). International deployment scope. Batch API calls are exempt from rate limits.", "limits": { "default": { "spend_threshold_usd": null, "rpm": 15000, "tpm": 5000000, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Qwen", "provider_id": "qwen", "model_id": "qwen-plus", "model_name": "Qwen-Plus", "docs_url": "https://www.alibabacloud.com/help/en/model-studio/rate-limit", "notes": "Limits are per Alibaba Cloud root account (aggregated across all API keys). International deployment scope. Batch API calls are exempt from rate limits.", "limits": { "default": { "spend_threshold_usd": null, "rpm": 600, "tpm": 1500000, "rpd": null, "tpd": null, "notes": null } } }, { "provider": "Qwen", "provider_id": "qwen", "model_id": "qwen-turbo", "model_name": "Qwen-Turbo", "docs_url": "https://www.alibabacloud.com/help/en/model-studio/rate-limit", "notes": "Limits are per Alibaba Cloud root account (aggregated across all API keys). International deployment scope. Batch API calls are exempt from rate limits.", "limits": { "default": { "spend_threshold_usd": null, "rpm": 600, "tpm": 5000000, "rpd": null, "tpd": null, "notes": null } } } ] }