Free AI APIs in 2026: 8 options and their real limits

Free AI APIs compared

The table below compares every free LLM API we checked on October 29, 2026 on each provider's own pricing, limits and docs pages. Providers change these caps often, so check your account dashboard before you build against a specific number.

• Gemini's free tier has the strongest free models, though Google uses your free-tier content to improve its products.
• OpenRouter gives you the most variety, but 10 of its 16 free endpoints may train on your prompts.
• Cerebras is a 30-day trial that requires a card.

Google Gemini API free tier

The Gemini API free tier gives you free input and output tokens on Gemini 3.8 Flash (gemini-3.8-flash), Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Gemini 3.1 Flash-Lite, Gemini 3 Flash Preview, Gemini 2.5 Pro, Gemini 2.5 Flash and Gemini 2.5 Flash-Lite, plus Gemini Embedding and several text-to-speech models. You don't get Gemini 4 Argon, Gemini Omni Flash or the Nano Banana image models on the free tier, and Grounding with Google Search is turned off for Gemini 3.8 Flash unless you pay. On the paid tier, Gemini 3.8 Flash costs $0.75 per 1M input tokens and $3.75 per 1M output tokens through December 31, 2026, before jumping to $1.50 and $7.50 on January 1, 2027.

Rate limits apply per project rather than per Gemini API key, and daily request quotas reset at midnight Pacific time. Google stopped listing per-model free numbers in its public docs, so you have to open Google AI Studio to see your project's actual limits (and preview or experimental models get tighter caps). Anything you send on the free tier is used to improve Google's products, while paid-tier traffic isn't.

OpenRouter free models

OpenRouter has 16 OpenRouter free models with IDs ending in :free, including google/gemma-4-31b-it:free, nvidia/nemotron-3-ultra-550b-a55b:free and thinkingmachines/inkling:free. If you don't care which one answers, you can point your requests at the openrouter/free router and let it pick an available free model at random.

Free models are capped at 20 requests a minute, and the 50 requests a day are shared across all :free models; the daily cap jumps to 1,000 once you've bought at least $10 of credits in total. Spinning up extra API keys or separate accounts won't raise that cap. Keep an eye on privacy here too: OpenRouter tags 10 of the 16 free endpoints as possibly training on prompts, including the NVIDIA, Thinking Machines, Poolside and Liquid ones.

Groq, Mistral, Cloudflare and Cohere free tiers

• Groq's Free plan gives you 30 requests a minute, 1,000 requests a day, 8,000 tokens a minute and 200,000 tokens a day on openai/gpt-oss-120b, openai/gpt-oss-20b and qwen/qwen3.8-27b, plus 20 requests a minute and 2,000 a day on Whisper Large v3. You don't need a payment method unless you upgrade to the paid Developer tier.
• Mistral turns on API access by default in Mistral Studio's Free mode with no credit card required. You get a monthly usage allowance, and you can check your per-model rate limits on the Limits page of the admin panel.
• Cloudflare Workers AI includes 10,000 free Neurons a day on every account, which covers about 150 requests of 1,000 input and 500 output tokens on gpt-oss-120b (66 Neurons each) or about 460 of those requests on Llama 3.1 8B fp8-fast (22 Neurons each). Text generation models are rate limited to 300 requests per minute, unless the model requires the Workers Paid plan.
• Cohere creates a free trial key automatically with every new account, giving you 1,000 API calls a month and 20 requests a minute across chat models like Command A, Command A Reasoning, Command A Vision, Command R+, Command R, Command R7B and North Mini Code. Trial keys aren't allowed for production or commercial use.

Free tiers that ended or became trials

Four providers that often appear on free AI API lists now need a card or a purchase, or have shut down.

• Cerebras now gives new accounts $5 in trial credits that require a verified payment method, cap gpt-oss-120b and qwen-3.8-27b at 5 requests a minute, and expire after 30 days.
• GitHub Models was fully retired on July 30, 2026, taking its inference API, playground, model catalog and bring-your-own-key option offline.
• Hugging Face Inference Providers gives free accounts no monthly credits, while PRO accounts and Team or Enterprise seats get $2.00 a month.
• SambaNova Cloud's Free plan requires you to add a payment method and buy credits before you can send your first request.

Claude, GPT and Gemini 4 Argon without per-token billing

None of the free tiers above give you a free Claude API or a free OpenAI API for GPT-6 models, and Google keeps Gemini 4 Argon locked to paid billing on the Gemini API.

If you already pay for Claude Pro or Max, a ChatGPT plan, Google AI Pro or Ultra, or a Kimi membership, Plan2API Free ($0, no card) lets you call the models in that plan, such as claude/claude-opus-5-5, openai/gpt-6-astra, gemini/gemini-4-argon-high and kimi/k3, through one OpenAI-compatible p2a_ API key at https://api.plan2api.com/v1; it is one of several subscription-to-API tools. Plan2API Free connects 1 subscription and allows 50 chat requests a day (resetting at 00:00 UTC) at 1 request per second with bursts up to 20, while your subscription's own usage limits still apply and Plan2API never bills per token. The curl example below calls Gemini 4 Argon.

curl https://api.plan2api.com/v1/chat/completions \
  -H "Authorization: Bearer $PLAN2API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini/gemini-4-argon-high",
    "messages": [{"role": "user", "content": "Write a haiku about rate limits."}]
  }'