Models & routing
Auto Free API guide
Free requests, without a deposit or money hold. Subject to availability and account limits.
Get started
Check Auto Free availability, then create a key and save the key and recovery link. Set MINIROUTER_KEY locally and send this request.
curl https://api.minirouter.sh/v1/chat/completions \
-H "Authorization: Bearer $MINIROUTER_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"minirouter/auto:free","messages":[{"role":"user","content":"Explain an API gateway in one sentence."}],"max_tokens":256}'The response names the answering model in x-minirouter-model and reports zero in x-minirouter-cost-usd and usage.cost. You can also use the signed-in Auto Free playground.
Supported requests
Chat Completions (/v1/chat/completions), Responses (/v1/responses), and Messages (/v1/messages) support JSON and SSE for text requests. Unsupported features return an error.
Use max_output_tokens for Responses or max_tokens for Chat Completions and Messages. Add stream: true for SSE. Do not send routing overrides: models, provider, providerOptions, route, or fallbacks.
Auto Free selects a model that supports your request and is available to your account and key. For text, it tries Poolside, Ling VL, Ling Santé, Ling Fin, then Ling 3.1 Flash before dispatch. Name a model directly to select it exactly.
poolside/laguna-s-2.1-freeinclusionai/ling-3.0-flash-vl-freeinclusionai/ling-3.0-flash-sante-freeinclusionai/ling-3.0-flash-fin-freeinclusionai/ling-3.1-flash-free
These models share one account allowance across all keys. Free requests currently accept text. Ling VL image input is not available on the free tier yet. Free inference never switches to a paid model.
Account limits
| Limit | Allowance |
|---|---|
| Concurrent requests | 1 per account |
| Requests per minute | 5 per account |
| Requests per UTC day | 50 per account |
| Input + output tokens per UTC day | 100,000 per account |
| Estimated input per request | 16,000 tokens |
| Output per request | 2,048 tokens maximum; 1,024 by default |
Limits apply across all keys on an account, alongside existing account and key limits. In-progress requests temporarily count their estimated input and maximum output toward your token allowance. Once complete, the allowance adjusts to actual usage. Requests sent to a provider count toward your request quota even if they fail.
Daily allowances reset at 00:00 UTC. Creating another key does not increase them. /me.freeInference returns eligibility, remaining requests and tokens, resetsAt, and request caps.
Errors and usage
- 429 free_quota_exceeded: wait for
Retry-After. Raising a key limit or adding credits does not raise the free allowance. - 503 auto_unresolvable: Auto Free is temporarily unavailable. Retry later; the router will not switch to paid inference.
400 invalid_request: simplify unsupported features or remove routing overrides.
View your requests in Activity. Auto Free requests always cost you $0.
Auto Free, chat trial, and paid Auto
- Auto Free uses
minirouter/auto:freein the API or playground, with daily quotas. Adding credits never turns it into a paid request. - Chat trial is the separate three-message offer at /chat, subject to availability. It covers browser chat, not integration API requests.
- Paid Auto uses
minirouter/autoor the Smart, Cheap, Code, and Roleplay intents. It uses credits and normal monetary holds. Choose a paid model explicitly to use it.
Playground conversations clear on reload; only model and reasoning preferences persist. The separate /chat flow offers account-scoped history and explicit saving.
A listed model may still be unavailable for your request. Each request must fit your account limits, key permissions, supported features, and available capacity.