Claude Fable 5.2

AnthropicChat & reasoningper-tokensave 50% vs official100.00%

Claude Fable 5.2 for reasoning, coding and agent workflows. Requires adaptive thinking; choose reasoning_effort for OpenAI-compatible requests.

Price

$4.00 Input · $20.00 Output / 1M

Max

$3.60 in

$18.00 out

Endpoint

/chat

Delivery pattern

Direct request-response

Context window

200K

API documentation

Build and copy the request

Choose the options this model accepts. The code below updates as you choose.

Step-by-step in the docs

POSThttps://apiking.ai/v1/chat/completionsBearer API keyReturns directly

OptionsOnly options published for this model are shown.

reasoning_effort

API_KEY="${API_KEY:-YOUR_API_KEY}"

curl --fail-with-body -sS https://apiking.ai/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "Claude Fable 5.2",
  "messages": [
    {
      "role": "user",
      "content": "Say hello."
    }
  ],
  "stream": false,
  "reasoning_effort": "low"
}'

Parameter reference

model

stringrequired

Exact model identifier for this request.

claude-fable-5-2

messages

arrayrequired

Conversation messages in the endpoint format.

Response lifecycle

Successful response example for this endpoint.

Response

{
  "id": "chatcmpl_xxx",
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "Hello!"
      }
    }
  ]
}

Errors and production

Retry rules, idempotency, and the minimum production checklist.

400 / 422

Fix the request fields or unsupported option; do not retry unchanged.

401

Check that the API key is present, active, and sent as a Bearer token.

402

The account balance is insufficient.

409

The Idempotency-Key is processing or was reused with another body.

429

Back off according to Retry-After before retrying.

5xx

Retry with exponential backoff and keep the original Idempotency-Key.

Keep API keys on your server for public mobile and browser applications.
Set a unique Idempotency-Key for every billable async operation.
Persist task_id so polling can resume after a restart.
Use a timeout for every HTTP request and surface the request ID in support reports.
Read live model metadata instead of hard-coding prices and supported options.

Price and tiers

save 50% vs official

List price, for everyone.

Input

$4.00/ 1M tokens

$8.00

Output

$20.00/ 1M tokens

$40.00

Cache read $0.1568 / 1M tokens

Cache write $5.40 / 1M tokens · 1h $8.60

The price shown here is exactly what you are billed — failed calls are refunded automatically.