- Free allowance
- No account and no API key.
- Verdict
- caveatusable, with a named limit or an unprobed read path
- Cost
- free tier
- Identity
- anonymousno card, no phone
- Binding limit
- request-rate60 requests per minute per IP address (rolling 60-second window); over the limit the service answers 429 rate_limit_exceeded with a retry-after header
- Evidence
- documented9 sources; a first-party page was read and the sentence quoted
- Freshness
- stale on 2026-11-29checked 2026-10-29 (live); stale on 2026-11-29
| Free tier | No account and no API key. Per IP address: 60 requests per minute, 20,000 requests per day, 10,000,000 tokens per day, at most 65,536 input tokens and 8,192 output tokens per request unit tokens; metered on request-count; the service chooses the model on every request; GET /v1/models lists one id, auto, and a requested model is ignored |
|---|
| Price | none stated |
|---|
| Limits | - request-rate (binding) 60 requests per minute per IP address (rolling 60-second window); over the limit the service answers 429 rate_limit_exceeded with a retry-after header source
- volume-quota 20,000 requests per day and 10,000,000 tokens per day per IP address, both reset at 00:00 UTC; 65,536 input tokens and 8,192 output tokens per request source
|
|---|
| Binding | request-rate the 60 requests per minute per IP address binds before the 20,000 per day |
|---|
| Identity | anonymous none: the guide lists Authentication as None; omit the Authorization header or send Bearer flm_anonymous, and any other key is rejected with 401 |
|---|
| Egress | not published |
|---|
| Interface | rest |
|---|
| Tooling | any OpenAI-compatible client; POST {base}/chat/completions with no Authorization header, or with Bearer flm_anonymous for SDKs that insist on a key. GET /limits returns the live limits as JSON and is not counted |
|---|
| Composes with | pairs with an agent loop that sets max_tokens; the model cannot be pinned, and the service picks one per request |
|---|
| Reliability | under load the service answers 429 rate_limited after making a request wait up to 8s for capacity, and 502 upstream_error when an answer fails mid-way; identical temperature-0 requests may be served from a short-lived cache |
|---|
| Notes | the requested model is never echoed because the service ignores it; the guide names no field that reports the model chosen, so which model answered a given request is unknown |
|---|
| Evidence note | documented the limits, the model list and the x-ratelimit headers were read by GET; no chat completion was sent, because probe issues GET and HEAD only, so whether a completion answers is… |
|---|
| Parsed numbers | - request-rate 60 requests / minute
- volume-quota 20000 requests / day
|
|---|