SSyncDev

Claude Fable 5.2

SyncDev model directory · Profile updated Nov 17, 2026

Compare with another model

Claude Fable 5.2 is Anthropic's top model for demanding reasoning and long-horizon agentic work, with a 1M context at $8/$40, 20% below Fable 5.1.

Key specifications

Specification score
70.6
Context window
1M
Max output
128K
Input / 1M tokens
From $8.00
Output / 1M tokens
From $40.00
License
Proprietary

The specification score summarizes recorded features and limits. It is not a benchmark result.

Pricing conditions: Indicative: $8 input / $40 output per 1M tokens; cache reads $0.20; cache writes $10 (5-min) / $16 (1-h); Batch $4 / $20; fast mode $16 / $80. Requires 30-day data retention. Confirm in your Anthropic console or cloud provider.

Inputs and outputs

Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.

Accepts

  • Text input
  • Image input

Produces

  • Text output

Capabilities

ReasoningFunction callingVision

What is Claude Fable 5.2?

Anthropic cut the price and tightened the rules in the same release. Claude Fable 5.2, launched on 9 November 2026, costs $8 per million input tokens and $40 per million output, down from $10 and $50 on Claude Fable 5.1. Output is reported to be up to 30% faster. But four behaviours that 5.1 tolerated now return a 400 error, and a team that swaps the model ID without reading the notes will find out in production.

Cheaper, faster, and the top of the range

This is Anthropic's model for demanding reasoning and long-horizon agentic work. It is not the default. Anthropic still tells most teams to start with Claude Opus 5.5, at half the price, and to move up only when evals on Opus at higher effort still fall short. That is a useful honesty: Fable 5.2 is the escalation tier, not the everyday driver.

Field Value
API ID claude-fable-5-2 (pinned snapshot)
Released 9 November 2026
Context window 1M tokens
Max output 128K tokens
Input / output Text and image in, text out
Thinking Adaptive, always on
Effort levels low, medium, high (default), xhigh, max
Knowledge cutoff June 2026
Retirement Not before 9 November 2027

Some third-party catalogues list PDF input. Anthropic's own pages say text and image. If you depend on documents, test a PDF request directly before assuming support either way.

Four changes that can break a 5.1 integration

  1. Prefix checks are enforced. If your code edits the system prompt, the tools or any earlier message, then sends back a thinking block from a later turn, the request fails with "The block is bound to a different conversation". Append-only histories are safe.
  2. Thinking blocks do not travel down. Fable 5.2 reads blocks from many Claude models, but nothing except Mythos 5.2 reads its blocks. Routing a conversation from Fable 5.2 to Opus 5.5 or Fable 5.1 silently drops the reasoning.
  3. The old computer-use tool is gone. computer_20251124 returns 400 everywhere except Amazon Bedrock; declare computer_toolset_20260801 instead.
  4. Advisor pairs are stricter. A Fable 5.2 executor accepts only Fable 5.2 or Mythos 5.2 as its advisor.

Two older rules still apply. Thinking cannot be switched off, and forced tool use (tool_choice of any or a named tool) is rejected. For schema-valid JSON, use auto with strict tool use or JSON outputs.

Effort: start at high

Anthropic's guidance is to begin at the default and climb only where you measure a gain. Most of the capability is already present at high. xhigh is for the hardest long-running coding and agent work; max is for ceiling tests. One behaviour change catches people out: a prompt line such as "answer directly without deliberating" no longer stops the model from thinking at medium and above. If easy turns cost too much, lower effort for those turns rather than arguing with the prompt.

It can also stop a long task early, leaving a list of what remains. That happens more at low and medium.

The bill in detail

Cache reads are $0.20 per million tokens, just 2.5% of base input, which rewards long, stable system prompts. Five-minute cache writes are $10 and one-hour writes $16. Batch halves the base to $4 and $20. Fast mode, enabled with speed: "fast" and no beta header, costs $16 and $80. All of it is indicative; check your console.

There is also a data condition. It requires 30-day retention and is not available under zero data retention unless Anthropic expressly authorises it. Regulated teams should clear that first.

Where it runs

It is on the Claude API for all customers, Claude Platform on AWS, both Google Cloud options, Microsoft Foundry on Anthropic infrastructure, and Amazon Bedrock as global.anthropic.claude-fable-5-2 or us.anthropic.claude-fable-5-2 on the bedrock-runtime endpoint. Server-side fallback, in beta, can retry refused requests on Claude Opus 5 or Opus 4.8.

Deciding between Fable and Opus

Take the ten tasks where Opus 5.5 at xhigh disappoints you. Run them on Fable 5.2 at high. If it solves several of them, route only that class of task up and keep the rest on Opus. If it solves none, the problem is likely the harness, not the model. More options sit on the Anthropic provider page.

How to evaluate Claude Fable 5.2

Compare the model on the work you need to complete.

Tasks and inputs

Claude Fable 5.2 is listed for Image-text-to-text, Text generation, Coding & Code Generation, Computer Use. It accepts text, image and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.

Capacity and cost

Documented limits: 1M context and 128K maximum output. The listed route has starting rates of $8.00 input and $40.00 output per million tokens. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.

Choosing a deployment

Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.

Frequently asked questions

$8 per million input tokens and $40 per million output tokens, with cache reads at $0.20 and Batch at $4/$20. Fast mode costs $16/$80.

These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.

Tencent

Hy4 Flash

Tencent Hy4 Flash is the cost-focused companion to Hy4, keeping vision and agent skills at a far lower token price. Availability is still unsettled.

1.05M context

Alibaba

Qwen4 Max

Qwen4 Max (qwen4-max) is Alibaba's first Qwen4-generation flagship: 1M-token context, text, image and video input, thinking on by default, $1.50/$4.50 per 1M.

1M context

Sarvam

Sarvam-105B-1711

Sarvam-105B-1711 is a refreshed open post-train of Sarvam's 105B MoE, tuned for tool calling, instruction following, agents and Indian languages.

— context

Alibaba

Qwen4 Max Preview

Qwen4 Max Preview was Alibaba's five-day early-access build of its Qwen4 flagship, offered only through a Token Plan subscription before GA qwen4-max.

1M context

Google

Gemini 4 Carbon

Gemini 4 Carbon is Google's strongest Gemini 4 reasoning model for long-horizon agentic coding, scientific research and cyber defence, with a 1M context.

1.05M context

Anthropic

Claude Mythos 5.2

Claude Mythos 5.2 is the verified-access edition of Claude Fable 5.2 for vetted cyber and life-sciences organisations, with a 1M context at $8/$40.

1M context