Is OpenRouter free to use?
Partly. OpenRouter's Free plan has no platform fee and gives “25+ free models” from 4 free providers at 50 requests a day. Every other model is paid from credits you buy, at the provider's price plus a 5.5% platform fee on pay-as-you-go.
A free model is its own entry, with an ID ending in :free, and OpenRouter warns that free variants “may have different rate limits or availability compared to paid versions”.
Source: OpenRouter pricing (openrouter.ai/pricing), OpenRouter docs: free variant (openrouter.ai/docs/guides/routing/model-variants/free)
What are the rate limits on OpenRouter's free models?
20 requests a minute and 50 a day, rising to 1,000 a day once the account has bought at least 10 credits in total. The tier is “selected by all-time credits purchased”, so the higher limit stays after those credits are spent.
Past a limit the API answers with a 429 error. OpenRouter's advice is to “purchase at least 10 credits to raise your daily limit, or switch to the paid variant of the model, which has no platform-level request cap”.
OpenRouter's free-model limits
- Under 10 credits bought: 20 requests a minute, 50 a day
- At least 10 credits bought: 20 requests a minute, 1,000 a day
Source: OpenRouter docs: limits (text version) (openrouter.ai/docs/api/reference/limits.md), OpenRouter docs: limits (openrouter.ai/docs/api/reference/limits)
Why do OpenRouter's free models return a 402 error?
Usually a negative balance. OpenRouter's limits page names a negative credit balance as a cause of 402 errors, “including for free models”, and says “Adding credits to put your balance above zero allows you to use those models again.”
Source: OpenRouter docs: limits (text version) (openrouter.ai/docs/api/reference/limits.md), OpenRouter docs: limits (openrouter.ai/docs/api/reference/limits)
Which models are free on OpenRouter?
The list changes as providers add and withdraw free endpoints. On the date above it held 14 models, among them Google's Gemma 4 31B and 26B, NVIDIA's Nemotron 3 Ultra 550B, Nemotron 3.5 Super VL and Nemotron 3.5 Lightning, Cohere's North Mini Code and Liquid's LFM3 1.7B.
To skip choosing, set the model to openrouter/free. The Free Models Router “automatically selects a free model at random from the available free models on OpenRouter”, keeping to models that support what the request needs, such as images, tool calling or structured outputs.
Source: OpenRouter model list (API) (openrouter.ai/api/v1/models), OpenRouter docs: Free Models Router (openrouter.ai/docs/guides/routing/routers/free-router)
Do OpenRouter's free models train on my prompts?
It depends on the provider serving the model. OpenRouter keeps “separate settings for paid and free models” for routing to providers that may train on prompts. Its provider table marks NVIDIA and Liquid “May train”, and on the date above each served its own free models alone.
Providers that may train “will not be routed to unless the model training toggle is switched on”, so with that setting off for free models, a free model only such a provider serves is not available to you. Turn it on only for prompts you are content to share.
Source: OpenRouter docs: provider logging (openrouter.ai/docs/guides/privacy/provider-logging), OpenRouter FAQ (openrouter.ai/docs/faq), OpenRouter endpoints for nvidia/nemotron-3-ultra-550b-a55b:free (openrouter.ai/api/v1/models/nvidia/nemotron-3-ultra-550b-a55b:free/endpoints), OpenRouter endpoints for liquid/lfm-3-1.7b:free (openrouter.ai/api/v1/models/liquid/lfm-3-1.7b:free/endpoints)
Can I use OpenRouter's free models in production?
OpenRouter advises against it: free models have low rate limits and “are usually not suitable for production use”. For production, call the paid variant of the same model, which “has no platform-level request cap”.
Source: OpenRouter FAQ (openrouter.ai/docs/faq), OpenRouter docs: limits (openrouter.ai/docs/api/reference/limits)