Awesome Free LLM APIs

Awesome Free LLM APIs is an open-source GitHub repository maintained by the mnfst organization that catalogs LLM API endpoints offering permanent free tiers — excluding time-limited trials and promotional credits. The list is licensed under CC0-1.0, making it freely usable and redistributable. As of its last push in May 2026, the repository had accumulated over 4,600 stars and 450 forks.

What It Is

This is a community-curated "awesome list" that aggregates LLM API providers and inference platforms offering no-cost access tiers for text inference. It is organized into two main categories: Provider APIs (run by the companies that train or fine-tune the models themselves) and Inference Providers (third-party platforms hosting open-weight models). Each entry includes the base URL, supported models, context window sizes, max output tokens, modality support, and rate limits. All endpoints are noted as OpenAI SDK-compatible unless explicitly stated otherwise.

What the List Covers

The repository documents a wide range of providers across geographies and model families:

• Provider APIs: AI21 Labs, Aion Labs, Alibaba Cloud Model Studio (Qwen), Cohere, DeepSeek, Google Gemini, Mistral AI, xAI (Grok), and Z AI (Zhipu AI / GLM)
• Inference Providers: Cerebras, Cloudflare Workers AI, GitHub Models, Groq, Hugging Face, Kilo Code, LLM7.io, ModelScope, Nebius, Nscale, NVIDIA NIM, Ollama Cloud, OpenRouter, OVHcloud AI Endpoints, and SiliconFlow
• Each entry links directly to the provider's API key page for quick onboarding

Notable entries include Google Gemini (with a footnote that the free tier is unavailable in EU/UK/Switzerland), Groq (ultra-fast LPU inference), Cerebras (ultra-fast inference at ~2,600 tokens/second), and OpenRouter (~28 free models with a :free suffix).

Scope and Contribution Rules

The list explicitly excludes trial credits and time-limited promotions — only permanent free tiers qualify. The contributing guide asks contributors to include the provider, endpoint, rate limits with a link to documentation, and a few notable models. This keeps the list actionable and verifiable rather than aspirational.

Architecture and Format

Each provider entry follows a consistent table format covering model name, context window, max output, modality (text, vision, audio, code, embeddings, reasoning), and rate limits. Footnotes are used to surface important caveats such as data-use policies, regional restrictions, and credit expiration details. The repository's primary language is listed as JavaScript, suggesting tooling or automation around the list itself.

Update: Active Development as of May 2026

The repository was created in March 2026 and last pushed in May 2026, indicating active curation. GitHub topics include ai-agents, llm-router, llm-routing, openai, gemini, ollama, openclaw, and openclaw-plugin, reflecting its positioning within the broader LLM tooling and routing ecosystem. The list already covers models released in 2025–2026 such as Gemini 2.5, Llama 4, Qwen3, and Grok 4 variants.