Google Gemini API
20 to 500 requests a day per model
Gemini 3.8 Flash, 3.6 Flash, 3.5 and 3.1 Flash-Lite, 2.5 Flash and Flash-Lite, text-to-speech and embedding models. Not free: Gemini 4 Argon, Gemini 3.1 Pro and the Nano Banana image models.
Flash models 5 requests a minute and 20 a day; Flash-Lite 15 a minute and 500 a day; embeddings 100 a minute and 1,000 a day. Limits are per project, not per key, and reset at midnight Pacific time.
No card neededOpenAI-compatible
Base URL https://generativelanguage.googleapis.com/v1beta/openai/
On the free tier Google uses your prompts to improve its products. Do not send private data.
How to get a Gemini API key for free, step by step →Source: ai.google.dev, read 24 October 2026
Gemma 4 on the Gemini API
14,400 requests a day
Gemma 4 31B and Gemma 4 26B A4B, input and output free of charge. There is no paid tier.
30 requests a minute and 16,000 tokens a minute, per model.
No card neededOpenAI-compatible
Base URL https://generativelanguage.googleapis.com/v1beta/openai/
Source: ai.google.dev, read 24 October 2026
Groq
1,000 requests a day
The free plan: gpt-oss-120b, gpt-oss-20b, Qwen3.8-27B and the Whisper speech-to-text models.
Chat models 30 requests a minute, 8,000 tokens a minute and 200,000 tokens a day. Whisper 20 requests a minute, 2,000 a day and 7,200 audio seconds an hour.
No card neededOpenAI-compatible
Base URL https://api.groq.com/openai/v1
How to get a Groq API key for free, step by step →Source: console.groq.com, read 25 October 2026
OpenRouter
50 requests a day
Every model whose ID ends in ":free", plus the openrouter/free router, which picks an available free model for each request.
20 requests a minute. The daily limit rises to 1,000 once an account has bought at least 10 credits.
No card neededOpenAI-compatible
Base URL https://openrouter.ai/api/v1
How to get an OpenRouter API key for free, step by step →Source: openrouter.ai, read 25 October 2026
Cloudflare Workers AI
10,000 Neurons a day
Included in the Workers Free plan: most text models, including gpt-oss-20b, gpt-oss-120b, Gemma 4 26B and Llama 3.3 70B. Some models (Kimi, GLM-5.x, DeepSeek V4) need Workers Paid.
Neurons are Cloudflare's unit of compute, billed at $0.011 per 1,000, so the free allowance is worth $0.11 a day at each model's listed price. Resets at 00:00 UTC.
Not confirmedOpenAI-compatible
How to get a Cloudflare Workers AI API key for free, step by step →Source: developers.cloudflare.com, read 25 October 2026
NVIDIA (build.nvidia.com)
No daily limit stated
Free access to NIM API endpoints for prototyping, through the NVIDIA Developer Program. An API key needs phone verification with a one-time code first.
The verification screen lists unlimited API requests without daily limits; the rate depends on the model and current traffic. Phone verification is offered in 231 locations, which in October 2026 do not include Pakistan, Bangladesh, Iran or Russia.
No card neededOpenAI-compatible: not confirmed
Source: developer.nvidia.com, read 25 October 2026
Cohere
1,000 API calls a month
Free trial API keys for every new account, on all Command models. Trial keys cannot be used for commercial purposes.
Every chat model (Command A, A+, A Reasoning, A Vision, A Translate, Command R, R+, R7B, North Mini Code) 20 requests a minute on a trial key; Embed 2,000 inputs a minute; Rerank 10 requests a minute.
No card neededOpenAI-compatible
Base URL https://api.cohere.ai/compatibility/v1
How to get a Cohere API key for free, step by step →Source: docs.cohere.com, read 25 October 2026