Rate limits

View your organization’s configured limits on the platform Limits page. If you need higher rate limits, reach out to support@subconscious.dev.

Advertised caps per model (API): GET /v1/models. 429 response headers are documented on each inference endpoint in the API reference.

Handling 429s

POST /v1/chat/completions, /v1/messages, /v1/responses.

Route	`retry_after_seconds` in JSON
Chat Completions, Responses	`error.details.retry_after_seconds`
Messages	Use `Retry-After` on JSON errors (body omits the field)

Retry-After is seconds until the limit counter resets (minimum 1). Rolling windows for minute/hour/day; calendar UTC month for month.

Window	`error.code`
minute, hour	`rate_limited`
day, month	`quota_exceeded`

403 entitlement_suspended — no Retry-After.

Example: Chat Completions / Responses (429)

Response headers

HTTP/1.1 429 Too Many Requests
Retry-After: 42
x-request-id: req_019c4f2a8b3d4e5f
request-id: req_019c4f2a8b3d4e5f
Content-Type: application/json

Response body

{
  "error": {
    "message": "request minute limit exceeded",
    "type": "rate_limit_error",
    "code": "rate_limited",
    "request_id": "req_019c4f2a8b3d4e5f",
    "details": {
      "retry_after_seconds": 42,
      "scope": "org:org_abc:subconscious/glm-5.3-marathon",
      "current_usage": 600
    }
  }
}

Request (normal inference auth — no extra rate-limit headers)

POST /v1/chat/completions HTTP/1.1
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

Example: Messages (429)

Response headers

HTTP/1.1 429 Too Many Requests
Retry-After: 42
x-request-id: req_019c4f2a8b3d4e5f
request-id: req_019c4f2a8b3d4e5f
Content-Type: application/json

Response body

{
  "type": "error",
  "error": {
    "type": "rate_limit_error",
    "message": "request minute limit exceeded"
  }
}

Request

POST /v1/messages HTTP/1.1
x-api-key: YOUR_API_KEY
anthropic-version: 2023-06-01
Content-Type: application/json

Streaming

Before any stream bytes: JSON error + headers as above. After bytes start: HTTP 200 and the wait hint only in the stream error event (no Retry-After on the HTTP response). Limit rejections are not metered as completions.

Get started

Features

Ways to use

Resources

Handling 429s

Example: Chat Completions / Responses (429)

Example: Messages (429)

Streaming

​Handling 429s

​Example: Chat Completions / Responses (429)

​Example: Messages (429)

​Streaming

Handling 429s

Example: Chat Completions / Responses (429)

Example: Messages (429)

Streaming