Best free models on OpenRouter
FreeOpenRouter lists dozens of models you can call at $0 (rate limits apply). These are the best free models on OpenRouter right now, grouped by what you want to do: general use, coding, chat and reasoning, vision, and uncensored.
As of June 2026, picking the best free model on OpenRouter comes down to what you are doing:
- General use: Owl Alpha for the biggest context and best all-round quality, or the Free Models Router to skip choosing
- Coding: Qwen3 Coder 480B for large repositories, or Kimi K2.6 for agentic edits
- Chat and reasoning: Nemotron 3 Ultra or Qwen3 Next 80B
- Vision: Gemma 4 31B for images and video
- Roleplay or unrestricted: Hermes 3 405B Instruct or Venice's Uncensored model
All of them have a $0 tier on OpenRouter. If you need always-on use with no rate limits, self-host an open-weight pick instead.
The best free model on OpenRouter right now is Owl Alpha, OpenRouter's own free stealth model: it pairs a 1M token context window with strong all-round quality at $0 (rate limits apply).
If you would rather not pick, the Free Models Router auto-routes each request to whatever free model is available, with zero setup. For coding, Qwen3 Coder 480B is the free specialist, and Kimi K2.6 is the strongest free open-weight model for real agent work.
Every model here costs $0 on OpenRouter, so your only constraint is each model's rate limits.
What most people should reach for first: strong quality and big context windows at $0, or a router that picks a free model for you.
- 11Owl AlphaOpenRouterOpenRouter's free stealth model: a 1M token context window and strong all-round quality at $0.Context1.049MInputFree
- 22Free Models RouterOpenRouterAuto-routes every request to whatever is free right now. The zero-setup pick.Context200KInputFree
- 33Kimi K2.6Moonshot AIThe strongest free open-weight model for real agent work (47.6% on CursorBench).Context262KInputFree
Why are some models with an input or output price, like Kimi K2.6, shown as free? OpenRouter publishes a separate $0 endpoint for several paid models. Kimi K2.6, for example, is free to call at openrouter.ai/moonshotai/kimi-k2.6:free (rate limits apply).
Free models built for code and agentic edits, including a 1M context specialist for large repositories.
- 11Qwen3 Coder 480B A35BQwenOpen-weight coding specialist with a 1M context window for large repositories.Context1.049MInputFree
- 2Context262KInputFree
- 33gpt-oss-120bOpenAIOpenAI's open-weight model, a capable free generalist for code and chat.Context131KInputFree
Fast, capable picks for general chat and step-by-step reasoning at no cost.
- 11Nemotron 3 UltraNvidiaNvidia's open-weight flagship with a 1M context window, free on OpenRouter.Context1MInputFree
- 22Qwen3 Next 80B A3B InstructQwenFast open-weight MoE that handles general chat and reasoning at $0.Context262KInputFree
Free multimodal models that read images (and video or audio) alongside text.
- 11Gemma 4 31BGoogle DeepMindGoogle's open-weight vision model: reads images and video, free to call.Context262KInputFree
- 22Nemotron 3 Nano OmniNvidiaNvidia's free multimodal nano model for text, image, and audio input.Context256KInputFree
Lightly aligned open-weight models for roleplay and unrestricted prompts. Run them yourself for no rate limits.
- 11Hermes 3 405B InstructNous ResearchNous Research's large open-weight model, lightly aligned and free to run.Context131KInputFree
- 2Context33KInputFree
What is the best free model on OpenRouter?
Owl Alpha is the best free model on OpenRouter right now. It is OpenRouter's own free stealth model, with a 1M token context window and strong all-round quality, at $0 per token (rate limits apply). If you would rather not pick a specific model, the Free Models Router auto-selects a free model for every request.
What is the best free coding model on OpenRouter?
Qwen3 Coder 480B is the best free coding model on OpenRouter, an open-weight coding specialist with a 1M token context window for large repositories. Kimi K2.6 is the strongest free model for agentic edits (47.6% on CursorBench), and Poolside's Laguna M.1 is free while in preview. See our best free coding models on OpenRouter ranking for the full list.
Are these OpenRouter models really free?
Yes. Every model on this list costs $0 per token on OpenRouter, so you can run them end to end without paying. The catch is rate limits: free models cap your requests, and stealth or preview models can change or go offline without notice. For always-on or production use, a cheap paid model or a self-hosted open-weight model is more dependable.
What is the best free model on OpenRouter for chat or roleplay?
For general chat, Nemotron 3 Ultra and Qwen3 Next 80B are strong free picks. For roleplay or unrestricted prompts, the lightly aligned open-weight models are a better fit: Hermes 3 405B Instruct from Nous Research, or Venice's Uncensored model. All run at $0 on OpenRouter.
What is the best free Qwen model on OpenRouter?
Qwen3 Coder 480B is the best free Qwen model on OpenRouter for code, with a 1M token context window and open weights. For general chat and reasoning, Qwen3 Next 80B A3B is the faster free Qwen pick. Both cost $0 per token.
Models
12Filter
FreeUpdated
June 2026