OpenRouter free models: limits, keys and failover

What is free on OpenRouter

Models with a free variant are used at no cost by adding the :free suffix to the model id; each free variant is its own catalog entry. The openrouter/free router picks a free model that supports your request. Paid usage draws on prepaid credits.

Examples listed as free when we checked: openrouter/free , google/gemma-4-31b-it:free , nvidia/nemotron-3-ultra-550b-a55b:free , google/gemma-4-26b-a4b-it:free . Check the current catalog and your account access. Ficelle refreshes catalog data and applies eligibility checks during selection.

Listed free model examples on OpenRouter

• Gemma 4 google/gemma-4-31b-it:free
• Nemotron 3 Ultra nvidia/nemotron-3-ultra-550b-a55b:free

OpenRouter free-tier limits

Free variants are limited to 20 requests per minute. The daily cap is 50 requests for accounts that have bought fewer than 10 credits in total, and 1,000 requests for accounts that have bought at least 10. OpenRouter says free models have low rate limits and are usually not suitable for production. A negative credit balance can return 402 errors even on free models, and upstream providers can return 429 when they are at capacity.

Get a OpenRouter API key

Create an OpenRouter account, then create a named key. Each key can carry its own credit limit.

https://openrouter.ai/keys

OpenAI compatibility

OpenRouter documents its API as a drop-in replacement for the OpenAI SDK.

https://openrouter.ai/api/v1

Data and privacy

OpenRouter logs basic request metadata (timestamps, model, token counts) and does not log prompts or completions unless you opt in. Upstream providers apply their own policies, and some may train on prompts; account settings control whether free and paid requests can reach them.

Through Ficelle, your OpenRouter key stays on your machine and prompts go straight to OpenRouter: there is no Ficelle proxy in between.

Before you rely on it

• A negative balance can block free models with a 402. Check the balance and error metadata before retrying. GET /api/v1/key reports free_model_daily_requests (used, limit, remaining) for the UTC day where the free-model cap applies; it does not report RPM. Limits documentation rechecked on 6 October 2026.
• The free catalog changes often: some :free variants carry an expiration date.
• The daily cap depends on credits bought over the account lifetime, not the current balance.
• Free variants can have different limits or availability than the paid versions.
• Upstream providers of free models may train on prompts unless you exclude them in settings.

Use OpenRouter with automatic failover

Ficelle can include OpenRouter behind your local OpenAI-compatible endpoint when the integration is enabled and its models pass free-access and capability checks. OpenRouter is a Core reference provider. Configure access and confirm your account can use the selected models. Another eligible candidate must be available for failover to succeed. Local routing guards do not control upstream billing when an account permits paid usage beyond its free allocation. See strict-zero routing and account limits.

ficelle set-key openrouter      # hidden prompt, stored locally
ficelle doctor --text          # check provider configuration; account access is separate

from openai import OpenAI

client = OpenAI(base_url="http://127.0.0.1:8646/v1", api_key="ficelle-local")

# Ficelle selects among configured, enabled, eligible models.
# If none can serve the request, it returns an error.
client.chat.completions.create(
    model="ficelle/auto-tools",
    messages=[{"role": "user", "content": "..."}],
)

Pick a model type

Choose a model type for your request. Eligibility depends on the installed provider configuration, catalog data, free-access evidence and capability checks. A catalog listing alone does not confirm that your account can call a model.

• Tool calling
• JSON & structured output
• Fast, low-latency
• Reasoning
• Coding assistants
• Agent orchestration
• Long context
• Context compression
• Vision
• Audio input
• Video input
• Multimodal

Questions