None
None
Long Story Video Skill
Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.
Ads Video Skill
Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.
3D Science Explainer Video Skill
Convert scientific concepts into stunning 3D explain animations
Feedback
freeTrialImage.bannerPity
freeTrialImage.upgradeUnlock
- ✓freeTrialImage.benefitHd
- ✓freeTrialImage.benefitWatermark
- ✓freeTrialImage.benefitUnlimited
claude fable
Need deeper reasoning than Opus 5.5 can give? Claude Fable 5.2 offers a million-token context and effort-scaled thinking for long agent jobs.
All Tools
Discover our comprehensive AI-powered animation toolkit
Muse Video
Seedance2.5
The Future of AI Video Is Here.
Seedance 2.0
The Future of AI Video Is Here.
Veo3.1
Create Stunning Videos with Veo3.1
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
Seedance 2.0
The Future of AI Video Is Here.
Gemini Omni 1.2 Flash
What Claude Fable 5.2 Brings to Hard Problems
When Opus stops being enough, Claude Fable 5.2 picks up the hard reasoning and the long agent runs — on the same spec sheet as Claude Mythos 5.2.
- Engineered for Heavy Reasoning LoadsIf pushing effort to the top on Claude Opus 5.5 still misses your bar, this is the next tier up for stubborn reasoning problems and agent tasks that run for hours.
- A Million Tokens In, 128k OutDefault and maximum context both sit at one million tokens for Claude Fable 5.2 and Claude Mythos 5.2, and either model can write responses of up to 128k tokens.
- Thinking Depth Set by EffortThere is no off switch for thinking — the effort parameter dials its depth from low through medium, high, xhigh, and max, and the API starts you at high.
Putting Claude Fable 5.2 Into Production
Settle the build path, justify the upgrade with eval results, and close out every breaking change before real traffic arrives.
Claude Fable 5.2 Feature Breakdown
On-demand compaction, tools that change mid-conversation, the computer use toolset, fast mode, strict tool use, and how refusals are handled.
On-Demand Compaction (Beta)
With the compact-2026-09-04 beta header you decide exactly when older turns are summarized, keeping your history append-only so the prefix check passes.
Swap Tools Mid-Conversation (Beta)
Rather than rewriting the tools array, add or drop tools inside a system message and send the inline-tools-2026-09-15 header on the Claude API, Claude Platform on AWS, and Anthropic-managed Google Cloud.
Computer Use Through the Toolset
Access computer use by way of the computer_toolset_20260801 toolset — declaring the legacy computer_20251124 tool now returns a 400 invalid_request_error everywhere except Amazon Bedrock.
Where Fast Mode Runs
You can use fast mode on the Claude API, on Anthropic-managed and Google-managed Google Cloud alike, and inside Claude in Microsoft Foundry where it ships as a separate model, claude-fable-5-2-fast.
Guaranteed Schemas With Strict Tool Use
To get schema-valid JSON, keep tool_choice on auto and set strict: true, or move to JSON outputs; forcing a tool raises an error, and customer-managed encryption keys are incompatible with structured outputs.
Refusals, Fallbacks, and What You Pay
Refusals arrive as HTTP 200 with stop_reason refusal and a stop_details object; certain categories may retry on Claude Opus 4.8 or Claude Opus 5, and the fallback credit refunds prompt-cache spend.
Claude Fable 5.2: Questions Answered
Straight answers on where to start, how migration works, and what Claude Fable 5.2 costs.
Do I begin on Claude Opus 5.5 or go straight to Claude Fable 5.2?
For most workloads, begin on Claude Opus 5.5. Move up to Claude Fable 5.2 only after higher-effort evals on Opus keep falling short of your target.
What causes the 400 error when I edit an earlier turn?
The prefix check is responsible: it inspects the system prompt, tools, and any earlier messages ahead of a thinking block, and a single change fails the call. As a stopgap, send the thinking-binding-controls-2026-08-01 header with prefix_mismatch_behavior set to drop_block.
Who can read thinking blocks produced by Claude Fable 5.2?
Just Claude Fable 5.2 and Claude Mythos 5.2 can read one another's blocks. Whatever the active model cannot read is stripped before delivery, and stripped blocks are not billed.
Is there a way to disable thinking or force one tool?
Not currently. Sending thinking enabled with budget_tokens or thinking disabled both return 400, as does setting tool_choice to any or to a named tool. Use effort levels and strict tool use instead.
How do Claude Fable 5.2 prices stack up against Claude Fable 5.1?
Input and output both cost less than on Claude Fable 5.1, which was priced at $10 per million input and $50 per million output tokens. Cache reads hold at 0.025 times base input, and batch runs come in at $4 per million input tokens.
What should I expect if a request gets refused?
The response is HTTP 200 with stop_reason refusal and a stop_details object that names the policy area involved. Some categories qualify for a server-side fallback retry, and what you are charged depends on the category.
Try Claude Fable 5.2 Before You Commit
Establish the need before anything moves: confirm whether Claude Opus 5.5 already meets your bar, lock in a build path, and rehearse all four breaking changes inside your own integration ahead of shifting production traffic.