generated: '2026-07-19'
method: searched
source: https://docs.kalpalabs.ai/rate-limits-and-errors
model: per-key token bucket
description: >-
  Each API key has a sustained requests-per-minute allowance and a burst capacity
  (a token bucket: `burst` requests available at once, refilling at the sustained
  rate). Exact values are set per key at provisioning; raise them via hello@kalpalabs.ai.
signaling_headers:
- header: X-RateLimit-Limit
  meaning: Your sustained requests/minute
- header: X-RateLimit-Remaining
  meaning: Requests left in the bucket right now
- header: X-RateLimit-Reset
  meaning: Seconds until the bucket refills
- header: Retry-After
  meaning: Seconds to wait after a 429
over_limit:
  status: 429
  type: rate_limit_exceeded
  guidance: Honor Retry-After and back off; all generation requests are safe to retry.
rate_limits:
- name: sustained-rpm
  scope: per-key
  limit_count: null
  window: minute
  note: Provisioned per key; not a fixed public value.
- name: burst
  scope: per-key
  limit_count: null
  window: instantaneous
  note: Token-bucket burst capacity, provisioned per key.
request_caps:
- cap: text-per-request-or-turn
  limit_count: 8000
  unit: characters
- cap: turns-per-conversation
  limit_count: 64
  unit: turns
- cap: audio-per-turn
  limit_count: 25
  unit: MiB decoded WAV
caps_served_at: GET /v1/info (under `limits`)