Claude Fable 5.2
Anthropic's most capable public model for long, hard technical work, coding agents and research.
What is Claude Fable 5.2?
Claude Fable 5.2 is Anthropic's most capable generally available model, released on November 9, 2026. It is built for long, open-ended technical work: large codebases, multi-agent research and computer use, and it can coordinate many subagents on one project.
Fable 5.2 replaces Fable 5.1 and launched alongside Claude Mythos 5.2, which shares the same underlying model but has looser cybersecurity and biology safeguards and is limited to vetted organisations.
Anthropic positions Fable 5.2 as a lead agent: it plans a big task, hands parts to subagents and checks the results. On Anthropic's benchmarks it beats Claude Opus 5.5 on agentic coding (FrontierSWE v2: 69.9% vs 62.3%), computer use (OSWorld 2.1: 89.3%) and Humanity's Last Exam with tools (74.2%).
It is about 30% faster than the 5.1 models, 20% cheaper, and refuses harmless requests about 55% less often. A fast mode runs up to 2.5x quicker at double the price.
For everyday writing and coding, Anthropic still recommends Opus 5.5, which is cheaper. Fable 5.2 is worth it for multi-day, hard problems.
Key features of Claude Fable 5.2
- Multi-agent orchestration: Acts as a lead agent coordinating large numbers of subagents.
- Agentic coding: Works across large codebases and multiple repositories and tests its own changes.
- Computer use: Operates desktop apps and browsers, scoring 89.3% on OSWorld 2.1.
- Vision and documents: Reads images, charts and PDFs and builds 3D scenes and simulations.
- Fast mode: Up to 2.5x faster output for time-sensitive work.
- Wide availability: Claude apps, Claude Platform API, AWS, Google Cloud and Microsoft Foundry.
What you can use Claude Fable 5.2 for
- Long software engineering projects
- Research that needs many agents in parallel
- Automating desktop and browser workflows
- Analysing large reports, PDFs and data
- Hardware, robotics and scientific work
Pros and considerations
Strengths
- Top scores on agentic coding and computer use benchmarks
- Cheaper and faster than Fable 5.1
- Fewer false refusals
- Available on all major clouds
Things to check
- Much more expensive than Opus 5.5 or smaller models
- Risky cyber and biology requests may be routed to weaker models
- 30-day data retention by default for safety monitoring
- Overkill for simple writing or social media tasks
Claude Fable 5.2 pricing
Paid. Included in paid Claude plans; on the API it costs $8 / $40 per million input / output tokens at launch, so check the official site for current pricing.
Who Claude Fable 5.2 is best for
Teams tackling long, difficult engineering or research work that needs a top frontier model.
Independent listing by AI Agent Help, last reviewed October 2026. Features and pricing change; check the official website for the latest details. Spot something out of date? Tell us.
Frequently asked questions
When was Claude Fable 5.2 released?
November 9, 2026.
How much does Claude Fable 5.2 cost?
It is included in paid Claude plans. On the API it launched at $8 per million input tokens and $40 per million output tokens.
What is the difference between Fable 5.2 and Mythos 5.2?
Same base model; Mythos 5.2 has looser cyber and biology safeguards and is only for vetted organisations.
Where can I read more?
Read our full Claude Fable 5.2 guide on the AI Agent Help blog.