Credit costs
These weights are documented here as the canonical source. Clients should treat the table above as static — the previous
credits.weights field on GET /v1/usage has been removed in favor of a fixed shape that matches the iOS UsageDTO.
Tier allowances
Each tier sets caps on questions, text-to-speech minutes, and credits per monthly period.
Periods roll over at the first of each calendar month UTC. The exact rollover timestamp is returned as
period.resets_at on GET /v1/usage.
An explicit non-zero
credits_total on the user’s profile (for example, an early-access lifetime grant) overrides the tier’s credits cap. Question and TTS caps always come from the tier table.Checking your usage
Your current credit usage is included in your profile and bootstrap responses, and is also available as a dedicatedGET /v1/usage endpoint that returns the current period, tier with caps, used counters, and remaining allowance in one shot. See Billing and usage for the full response schema.
- Usage endpoint
- Profile endpoint
- Bootstrap endpoint
used.credits equals tier.caps.credits (or credits_used equals credits_total on the profile), the next AI operation that requires credits returns 429 Too Many Requests. Free-tier accounts recover at the start of the next monthly period.
The
questions and tts_seconds counters on GET /v1/usage currently return 0 until the usage aggregator is wired. The caps and remaining math are still authoritative — clients can rely on the shape today and pick up real counts when the aggregator ships.Rate limits
Credits and rate limits are separate mechanisms. Rate limits cap how many requests you can send per minute, regardless of your credit balance.
When a rate limit is hit, the response includes a
Retry-After header indicating when you can try again.
Rate-limited research requests with a valid session token may be captured to a retry queue rather than dropped entirely. Each account can have up to 5 pending queued requests.