Claude Fable 5.2 API pricing and how to call it

Price per million tokens

Fable 5.2 is 20% cheaper per token than Fable 5.1, which Anthropic lists at $10 and $50. On KeelRouter you pay $2.50 per million input tokens and $12 per million output tokens, about 30% of Anthropic's list price. Thinking tokens are billed as output.

USD per 1M tokens. Anthropic prices from its Claude Fable 5.2 model page, checked 2026-11-11. KeelRouter bills cache reads at 10% of its input price but never above Anthropic's cache read price, bills cache writes at 125% of its input price, and does not sell fast mode or the Batch API.

The effort setting decides the bill

Adaptive thinking is always on in Fable 5.2 and cannot be turned off. You steer how much it thinks with output_config.effort: low, medium, high, xhigh or max. The API default is high; Claude apps and Claude Code start at medium. Anthropic recommends starting at high, moving to xhigh only for the hardest long-running coding and agent work, and keeping max for measuring the ceiling.

The per-token price is the same at every level, but the number of tokens is not. On the Artificial Analysis Intelligence Index, an average task cost $3.10 at low, $3.80 at medium, $9.08 at high, $14.83 at xhigh and $26.54 at max at list prices, for scores of 54, 56, 62, 62 and 65. Going from high to max nearly tripled the cost for three points.

Thinking counts toward max_tokens. At high and above, set a large max_tokens (it is a hard cap on thinking plus answer), or the answer can come back empty with finish_reason "length".

How to call claude-fable-5-2

Use the model ID claude-fable-5-2 with one KeelRouter key. POST /v1/chat/completions takes the OpenAI format; POST /v1/messages takes the Anthropic format, so the official Anthropic SDKs work with only the base URL and key changed.

curl https://api.keelrouter.com/v1/chat/completions \
  -H "Authorization: Bearer $KEELROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-fable-5-2",
    "max_tokens": 32000,
    "output_config": {"effort": "medium"},
    "messages": [{"role": "user", "content": "Review this migration plan and list the risks."}]
  }'

import anthropic

client = anthropic.Anthropic(base_url="https://api.keelrouter.com", api_key="sk_...")
message = client.messages.create(
    model="claude-fable-5-2",
    max_tokens=32000,
    output_config={"effort": "medium"},
    messages=[{"role": "user", "content": "Review this migration plan and list the risks."}],
)
print(message.content[-1].text)

Limits to know before you switch

1M-token context window and 128K max output per request. Input is text and images; output is text. Reliable knowledge cutoff: June 2026.

Forced tool use (tool_choice "any" or "tool") returns a 400 error, and so do temperature, top_p and top_k: leave them out.

Anthropic requires 30-day data retention for Fable 5.2; it is not available under zero data retention unless Anthropic authorizes it.

Safety classifiers block high-risk cybersecurity and most research-biology requests. Anthropic says Fable 5.2 blocks about 55% fewer requests than Fable 5.1 in its early testing.

Claude models run on a Claude Code–compatible route. The provider adds a short system prompt (about 30 tokens) before yours; it counts toward input tokens, and answers can differ slightly from Anthropic’s own API.

FAQ

How much does the Claude Fable 5.2 API cost?

Anthropic lists it at $8 per million input tokens and $40 per million output tokens. On KeelRouter it is $2.50 and $12, pay as you go from a $5 balance.

Is Fable 5.2 more expensive than Opus 5.5?

Per token, yes: twice the price at Anthropic ($8 / $40 against $4 / $20), and about twice on KeelRouter ($12 against $6.10 per million output tokens). Per finished task it depends on the effort level; see our Fable 5.2 vs Opus 5.5 comparison.

What is the model ID?

claude-fable-5-2, the same as on the Claude API. Amazon Bedrock uses anthropic.claude-fable-5-2.

Can I turn thinking off to save tokens?

No. Adaptive thinking is always on for Fable 5.2. Lower output_config.effort to medium or low to make it think less.

More guides

• GPT Image 2.6 API pricing
• GPT Image 2.6 prompts
• Make a photo sing with an API
• AI face swap with lip sync: 8 ways we tried on the White Chicks car scene
• Claude Fable 5.2 prompting guide and community templates
• Gemini 4 Carbon API pricing and prompting guide