Claude Fable 5.1 API

Price check: official, OpenRouter and Wokey

USD per million tokens. Official and OpenRouter rates were checked by hand on 2026-09-29, excluding tax; the Wokey rate is read from the live price list.

Checked 2026-09-29 · Official pricing page · OpenRouter endpoint list

Use it in Claude Code

This model supports /v1/messages on Wokey. Set these variables, then start claude.

export ANTHROPIC_BASE_URL="https://api.wokey.ai"
export ANTHROPIC_AUTH_TOKEN="$WOKEY_API_KEY"
export ANTHROPIC_MODEL="claude-fable-5-1"

Claude Fable 5.1 at a glance

Release date

2026-09-01, alongside Mythos 5.1; the two are the same model with different safeguards

Positioning

For demanding reasoning and long-horizon agentic work. Anthropic says to start most workloads on Opus and use Fable 5.1 when your evals on Opus at higher effort still fall short

Context and output

1M context at standard pricing across the whole window; 128K max output; Anthropic lists its latency as “slower”

Input and cutoff

Text and image input; knowledge cutoff June 2026; same tokenizer as Fable 5, which uses about 30% more tokens than models before Opus 4.7

Thinking and effort

Adaptive thinking is always on and cannot be disabled; the API default effort is high (Opus 5.5 defaults to medium); effort can now change mid-conversation (beta) without invalidating the cache

Official price

$10 input / $50 output per million tokens, the same as Fable 5; cache read $0.25, 2.5% of input (Fable 5: $1.00); 5-minute cache write $12.50, 1-hour $20; Batch 50% off

Safeguards and data

About 60% fewer cyber safeguard interventions than Fable 5; refusals return stop_reason: "refusal" and can fall back to Opus 4.8 or Opus 5; 30-day retention, and zero data retention (ZDR) only with authorization

Model ID

claude-fable-5-1 (anthropic.claude-fable-5-1 on Bedrock; the same ID on Google Cloud, Microsoft Foundry and Claude Platform on AWS)

Checked 2026-09-29 · Sources: Anthropic · What’s new · Migration guide

Official benchmarks: Fable 5.1 vs Opus 5.5 vs Fable 5 vs Opus 5 vs GPT-6 Astra

From Anthropic’s Opus 5.5 launch page; these are vendor-reported results, and GPT-6 Astra scores are OpenAI-reported. The Fable 5.1 launch page uses older versions of several benchmarks and cannot be merged into this table, so the Fable 5 column is only filled where both pages report the same Fable 5.1 and Opus 5 scores; “—” means no comparable published score.

On the Opus 5.5 launch page Anthropic says the gap between Opus 5.5 and Fable 5.1 is narrower than these scores suggest. Fable 5.1’s launch testing ran with safeguards on, scoring blocked tasks as zero or handing them to Opus 4.8 or Opus 5, which Anthropic says likely lowered its scores. On the same launch page Mythos 5.1 scored 60.9% on Terminal-Bench 4.0 versus 55.8% for Fable 5.1.

Which to use: Fable 5.1, Opus 5.5 or Fable 5

vs Opus 5.5

Opus 5.5 lists at $4 / $20, two-fifths of Fable 5.1’s price, reads cache for less ($0.20 vs $0.25) and leads all nine benchmarks above. On Artificial Analysis’ current index Opus 5.5 scores 58 at max for $5.98 per task, and Fable 5.1 scores 53 for $7.63. Start on Opus 5.5 and try Fable 5.1 only when your own evals still fall short with Opus 5.5 at xhigh or max.

vs Fable 5

Same price with cache reads at a quarter of the cost, so there is no reason to stay on Fable 5. Anthropic estimates about 25% savings on typical work and up to about 45% on highly agentic work. Terminal-Bench 4.0 rose from 42.0% to 55.8% and Terminal-Bench-Science doubled from 24.7% to 52.6%. Check the migration list below first: forced tool use returns 400, and replaying thinking after editing history can too.

Choosing effort

The API default is high, and Anthropic’s migration checklist says to re-tune starting there rather than carry over Fable 5 settings. Artificial Analysis measured an 11x spread in output tokens across effort levels; on its current index high scores 51 at $3.91 per task and max 53 at $7.63, nearly double the cost for 2 points. Start max_tokens at 64K for xhigh and max, and use the new mid-conversation effort change (beta) to switch levels without invalidating the cache.

Effort documentation

Third-party and community reviews

• Artificial Analysis: At launch it scored 66 on the Intelligence Index at max, #1 at the time and 4 points above Fable 5; a task cost $3.76, 20% more than Fable 5, driven by about 1.7x the output tokens, while the cache price cut saved about $1.40 per task. Refusal fallbacks produced about 4% of output tokens. On the current v4.3.2 index it scores 53 at max (#5) for $7.63 per task and 51 at high for $3.91. Every Wokey meter is 24% of the official rate, so the same task comes to about $1.82 at max and $0.93 at high.
• Early-access customers (Anthropic launch page): Cognition: moved Opus 5 traffic in Devin to Fable 5.1 on launch day. Browserbase: 82% of tasks completed versus 74% for Opus 5 and 57% for Fable 5. Glean: judges preferred it about 2 to 1 over Fable 5. Every: about twice as fast as Opus 5 with half the tokens. Rogo: 20% fewer tokens. These are customer statements quoted by the vendor.
• Hacker News (1,419-point thread): An Anthropic employee in the thread says the writing style is more natural and the Terminal-Bench-Science score more than doubled over Fable 5. Most of the discussion complains that Claude models write densely or emptily, and some suspect the verbosity drives up token use; commenters who want terser answers point to GPT-5.6 Sol.

Migrating from Fable 5 or Opus 5: requests that return 400

These changes come from Anthropic’s What’s new page and migration guide. Check your requests, conversation history and response parsing before switching the model to claude-fable-5-1.

• tool_choice of any or tool returns 400 (“tool_choice: type "tool" and "any" are not supported for this model.”), including on count_tokens, so requests that worked on Fable 5 fail outright. Use auto with strict: true, structured outputs or an explicit instruction in the prompt. Through Wokey the gateway returns the same 400 locally without spending upstream quota.
• Thinking blocks work one way: Fable 5.1 reads earlier models’ thinking blocks, but earlier models cannot read Fable 5.1’s. After a fallback they are dropped silently and not billed; with the thinking-binding-controls-2026-08-01 beta header they are listed in input_transformations.
• Editing earlier turns (system, tools or messages) invalidates later thinking blocks. Accounts created on or after 2026-08-31 get a 400 “The block is bound to a different conversation”; set thinking.block_binding.prefix_mismatch_behavior: "drop_block" to drop them instead. Keep history append-only; server-side compaction, context editing, moving cache_control and changing effort are all safe.
• From Opus 5: thinking: {"type": "disabled"} and {"type": "enabled", "budget_tokens": N} both return 400 (as they have since Fable 5); remove the thinking field and control depth with effort. Through Wokey the gateway strips these thinking fields before forwarding, so the request does not fail, but thinking still runs and bills as output tokens.
• max_tokens covers thinking plus text and the default effort is high; re-size requests that disabled thinking on Opus 5, and start at 64K for xhigh or max. Responses can start with thinking blocks, so read text blocks by type; thinking.display defaults to "omitted".
• Limits carried over from Fable 5 still apply: a trailing assistant prefill and non-default temperature, top_p or top_k return 400, and the minimum cacheable prompt is 512 tokens.
• Behavior changes: parallel tool calls are less consistent (ask for batching in the system prompt); there are fewer progress notes between tool calls, so set display: "updates" (beta); small edits may rewrite whole files; output uses less formatting and denser prose.
• Refusals return stop_reason: "refusal" with stop_details, and fallbacks: "default" (beta) retries on Opus 4.8 or Opus 5; since 2026-09-24, pre-output refusals in low-false-positive categories are billed. Coming from Opus 5, note that Fable 5.1’s safety classifiers cover more categories and zero data retention needs authorization.

Request body after migration (/v1/messages)

{
  "model": "claude-fable-5-1",
  "max_tokens": 64000,
  "thinking": { "type": "adaptive", "display": "summarized" },
  "output_config": { "effort": "high" },
  "messages": [{ "role": "user", "content": "Find and fix the flaky test in this repo" }]
}

Catalog pricing snapshot

Prices are per 1M tokens; request-time pricing controls settlement.

Official price source: https://platform.claude.com/docs/en/models/fable-5-1/overview

Model capabilities

Context

1,000,000 tokens

Max output

128,000 tokens

Streaming

Supported

Tools

Supported

Vision

Supported

Supported API forms

/v1/messages

Send a request

/v1/messages

curl https://api.wokey.ai/v1/messages \
  -H "x-api-key: $WOKEY_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-fable-5-1","max_tokens":1024,"messages":[{"role":"user","content":"Hello"}]}'

Use Claude Fable 5.1 in your tools

The gateway is https://api.wokey.ai (OpenAI-compatible clients usually take https://api.wokey.ai/v1), and the model ID is claude-fable-5-1. Open the guide for the client you use.

• Use Claude Fable 5.1 in Claude Code: Anthropic Messages · base URL and API key setup
• Use Claude Fable 5.1 in Claude Desktop: Anthropic Messages · base URL and API key setup
• Claude API cost calculator: what Claude Fable 5.1 costs per month: Compare monthly cost on the Anthropic API, OpenRouter and Wokey for your usage and cache hit rate.
• OpenRouter alternative: Wokey vs OpenRouter: Compare token prices, payment fees, API forms, and upstream sources.
• Verifiable AI API: check a Claude Fable 5.1 response came from the official upstream: Which responses carry a tee.proof, and what a proof does and does not show.
• Verify a Claude Fable 5.1 response online: AI API relay verifier: Paste a response with its tee.proof and check it locally in your browser; nothing is uploaded.

Example production call

These usage values come from a completed Claude Fable 5.1 call that succeeded on its first attempt and was reconciled with its bill.

Total input

619529

Output

1342

Cached share of input

99.8%

Call duration

14.98 s

Usage field example

Compiled from this call’s billing record. input_tokens is uncached input; cache reads and writes are separate.

{
  "model": "claude-fable-5-1",
  "usage": {
    "input_tokens": 2,
    "output_tokens": 1342,
    "cache_read_input_tokens": 618475,
    "cache_creation_input_tokens": 1052
  }
}

Token accounting

These are the final per-million-token rates recorded for this call.

Billed amount

$0.054024

Cost = sum of tokens × their rate ÷ 1,000,000, rounded to six decimals after summing.

Source of this call

The transport record points to api.anthropic.com with OAuth authorization. Usage was metered upstream and the request returned HTTP 200.

Verify Prompt Cache

1. Keep the system, tools, and conversation prefix fixed and mark cache_control according to the cache rules.
2. Check cache_creation_input_tokens on the first response, then reuse the prefix and check cache_read_input_tokens on the next.
3. Reconcile uncached input, cache reads, cache writes, and output separately. Total input is the sum of the three input buckets.

Frequently asked questions

Related models and documentation

• Claude Fable 5
• Claude Opus 5
• Claude Opus 5.5
• Claude Opus 4.8
• API documentation
• All models and pricing
• Provider Node