Highest Free API Rate Limits
Free AI APIs ranked by the highest published rate limit of any model they serve free: requests per minute first, then requests per day.
A free tier is only as useful as its rate limit: a model you can call twice a minute will not carry a batch job, a demo with real users or an agent that makes ten calls per task. This chart ranks the verified free tiers and signup credits by the highest free limit the provider publishes for any one model, and the line under each value names that model. Embedding and rerank APIs are in it too; they tend to allow far more calls than chat models. To check whether a limit covers your own workload, put it into the rate limit calculator; to compare chat models one by one, see free LLM APIs.
| # | Offer | Best free limit | Card | The catch |
|---|---|---|---|---|
| 1 | Cloudflare Workers AI Free Tier | 3,000 requests/min@cf/baai/bge-m3 | no card | 10,000 Neurons is small (about 49K output tokens a day on Llama 3.3 70B at the published rate); Kimi, GLM-5 and DeepSeek V4 models need Workers Paid or prepaid credits. |
| 2 | Voyage AI Free Embedding and Reranking Tokens | 2,000 requests/minvoyage-4-large | no card | Without a payment method you are held to 3 requests and 10K tokens a minute; the published 2,000 RPM limits need a card on file (Usage Tier 1). Older model generations get no free tokens. |
| 3 | Alibaba Cloud Model Studio Free Token Quota | 600 requests/minqwen-plus | no card | Quota expires 90 days after activation (or model release/approval, whichever is latest), covers real-time inference only in the Singapore region, and is shared by the account plus its RAM users. |
| 4 | Jina AI 10M Free API Tokens | 500 requests/minjina-embeddings-v5-text-small | no card | The 10M tokens are one-time and shared across all Jina APIs; a web search costs from 10,000 tokens. The Reader models are CC-BY-NC 4.0, so self-hosting them commercially needs a license. |
| 5 | OVHcloud AI Endpoints Free Tier | 400 requests/minQwen3.8-27B | card | Needs a saved payment method (the US site also requires a $0.99 purchase); billing starts automatically once the credit is used up; anonymous API use is capped at 2 requests per minute. |
| 6 | Scaleway Generative APIs Free Tier | 300 requests/minglm-5.2 | no card | Free Tier applies to your most expensive tokens first; once exceeded, per-token billing starts with a 1,000-token minimum billing unit. |
| 7 | AI21 Labs Free Trial | 200 requests/minJamba Large | no card | Credit expires after three months or once exhausted; valid billing information is then required to continue using the AI21 platform. |
| 8 | Mixedbread Free Starter Credits | 100 requests/minToast 1 | no card | The $5 credits are one-time only with no monthly renewal; once spent you must add credits from billing settings or upgrade to Scale. |
| 9 | LLM7.io Free Tier | 60 requests/minDeepSeek-V4-Flash-0731 | no card | Token usage counts input plus output combined against the daily cap, and per-second/minute/hour request limits apply; a free token is required for the 100,000 tokens/day allowance. |
| 10 | Token Harbor Free Allowance | 60 requests/minDeepSeek V4.1 Flash | no card | The free allowance size is not published, it resets on a rolling 7-day period, and the free model lineup rotates. |
| 11 | Token Harbor Free Allowance | 60 requests/minMiMo V2.6 Flash | no card | The free allowance size is not published, it resets on a rolling 7-day period, and the free model lineup rotates. |
| 12 | NVIDIA NIM Free Inference Endpoints | 40 requests/mindeepseek-v4.1-flash, 10,000 requests/day | no card | Free endpoints are limited to catalog models tagged "Free Endpoint" and can be deprecated; you need a verified NVIDIA account and an API key, and no rate limits or SLA are published for the free tier. |
| 13 | Sarvam AI Free API Credits | 40 requests/minsarvam-105b | no card | Rate limits are enforced per account across all keys (60 req/min for most APIs on Starter), and the credits are denominated in INR. |
| 14 | Groq Free Tier | 30 requests/mingpt-oss-120b, 1,000 requests/day | no card | Rate limits apply at the organization level and are tight (e.g. 30 RPM and 1K RPD on many models); the docs caution limits may change and exceptions exist. |
| 15 | AMD Radeon Cloud Free APIs | 30 requests/minDeepSeek-V4-Flash | no card | AMD picks which models are served; per-key rate limits, a concurrency allowance, and a spend cap all apply, and only chat-completions and messages endpoints are served. |
| 16 | Requesty Free Models | 20 requests/minnemotron-3-ultra-550b-a55b, 200 requests/day | no card | Only the 13 designated zero-cost models are free; they are rate-limited per organization and free-model terms can change. |
| 17 | Routeway Free Models | 20 requests/minDeepSeek V4 Flash, 200 requests/day | no card | Free models run on a shared, low-priority queue with best-effort availability and one concurrent request — suitable for prototyping, with no uptime guarantee. |
| 18 | OpenRouter Free AI Models | 20 requests/minNemotron 3 Ultra, 50 requests/day | no card | Free models allow 20 requests/minute and 50/day until you have bought 10 credits (then 1,000/day). A negative balance blocks them, and the lineup rotates (24 variants listed, 17 of them chat models). |
| 19 | Cohere Free Trial API | 20 requests/mincommand-a-plus-05-2026 | no card | Trial API keys are rate limited and not permitted for production or commercial use. |
| 20 | NaraRouter Free Tier | 15 requests/minAgnes 2.5 Flash | no card | Free access covers only the three free models at 15 requests per minute under a fair-use policy; all other models require a paid plan. |
How this chart is ranked
Verified offers of type free tier, fully free, free credits or free trial (startup and student programs are left out because they need an application) where at least one model has a published free rate limit. Each offer is ranked by its best model: highest requests per minute (RPM) first; offers that publish only a daily cap follow, by requests per day (RPD). Ties go to offers that need no credit card, then by name. Token-per-minute caps are not compared.
Rows come from the same catalog as the offer pages, re-checked monthly. Values are what each provider states; aifree.dev never converts credits or tokens into dollars. See how we verify.
Frequently asked questions
- What do RPM and RPD mean?
- RPM is requests per minute and RPD is requests per day: how many API calls the free tier accepts in that window before it answers with a rate-limit error (HTTP 429). Many providers publish both, and the daily cap is usually the one you hit first.
- Does a high RPM mean I can make that many calls all day?
- No. A per-minute limit says how fast you may call, not how much. A daily cap, a token-per-minute cap or a one-time credit pool can end the free use long before the per-minute limit matters, so read the whole limits list on the offer page.
- Why does the chart name one model per provider?
- Limits are set per model, and big models usually get lower limits than small or embedding models. The row shows the model with the highest published limit; the offer page lists the limits of every model it serves.
- Where do these numbers come from?
- From the rate limits each provider publishes for its free tier, recorded per model on the offer and re-checked when the offer is verified. A provider that does not publish a number for a model is not ranked on that model; nothing is estimated.