The free-llm-api-resources repository is a curated directory of services offering free or trial-based access to large language model APIs. Maintained primarily by cheahjs with contributions from aliasfoxkde and DarkReaperBoy, the repository documents legitimate API providers that allow developers to experiment with LLM inference without upfront payment. The README is automatically generated from a Python script located at src/pull_available_models.py, ensuring the list stays current as provider offerings change.
The repository catalogs two main categories of services: those offering perpetual free access and those providing trial credits. Free providers listed include OpenRouter, which offers 20 requests per minute and 50 daily requests with various open models like Llama 3.3 70B and Hermes 3; Google AI Studio, providing access to Gemini and Gemma models with token-per-minute limits; NVIDIA NIM with 40 requests per minute; Mistral's La Plateforme and Codestral services; HuggingFace Inference Providers with monthly credits; Vercel AI Gateway; OpenCode Zen; Cerebras; Groq; Cohere; GitHub Models; Cloudflare Workers AI; and several others. Trial credit providers include Fireworks, Baseten, Nebius, Novita, AI21, Upstage, NLP Cloud, Alibaba Cloud, Modal, Inference.net, Hyperbolic, SambaNova Cloud, and Scaleway Generative APIs.
Each provider entry specifies rate limits, token quotas, and available models with direct links to their respective platforms. The documentation includes important caveats, such as Google AI Studio using data for training outside certain regions, Mistral requiring phone verification, and NVIDIA NIM having context window limitations. The repository explicitly excludes services that reverse-engineer existing chatbots or lack legitimacy, maintaining focus on official provider offerings.
GitGenius tracking data reveals the repository has grown modestly, gaining three stargazers between July 4, 2026 and the present check, reaching 24,987 total stars. The project maintains 316 subscribers. Issue and pull request response latency shows a median of 54.7 hours with a mean of 429.2 hours across 39 tracked items, indicating variable response times depending on issue complexity. The primary maintainer cheahjs has logged 28 events in the tracked period, substantially more than secondary contributors aliasfoxkde with 4 events and DarkReaperBoy with 3 events.
The repository's classification spans multiple domains including LLM APIs, Free APIs, Language Models, AI Resources, API Directory, Generative AI, Developer Tools, AI Services, Model Access, and API Endpoints. It serves as a practical resource for developers seeking to prototype with various language models without financial commitment, while the explicit warning against service abuse reflects awareness that free tier availability depends on responsible usage patterns. The repository's connection to other projects through overlapping contributors, including diegosouzapw/omniroute, anomalyco/opencode, and reflex-dev/reflex, suggests it occupies a recognized position within the broader AI development ecosystem.
