On 1 September 2026, Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1. In its announcement, the company described them as models for coding, knowledge work, and research.
Anthropic says the two products are the same underlying model with different safeguards. Fable 5.1 is generally available. Mythos 5.1 is limited to trusted-access programs.
What Anthropic announced
The company published the launch on its site as Introducing Claude Fable 5.1 and Claude Mythos 5.1. It also posted model documentation for Fable 5.1, with a release date of 1 September 2026.
Developers can call Fable 5.1 on the Claude API as claude-fable-5-1. Anthropic says the model is also available through Amazon Web Services, Google Cloud, and Microsoft cloud channels, including Azure and Microsoft Foundry.
Anthropic’s documentation lists a 1 million-token context window, 128,000-token maximum output, and a June 2026 knowledge cutoff. Input and output list prices stay at $10 and $50 per million tokens.
What Claude Fable 5.1 is
Claude Fable 5.1 is the generally available version. Anthropic positions it for long-running agentic coding, multistep research, and document, spreadsheet, and slide work.
The company says Fable 5.1 is available to Pro, Max, Team, and Enterprise users, as well as through the Claude Platform and cloud marketplaces.
Coding and knowledge-work claims
Anthropic reports higher scores than Claude Fable 5 on several of its published evaluations, including:
- Terminal-Bench-Science 0.1: 52.6% for Fable 5.1, compared with 24.7% for Fable 5
- Terminal-Bench 4.0: 55.8% for Fable 5.1
- CursorBench 3.2.0: 73.4% for Fable 5.1, compared with 70.5% for Fable 5
- GDPval-AA v2 knowledge-work score: 1853, compared with 1723 for Fable 5
Those figures come from Anthropic’s announcement and should be read as company-reported results, not as independent tests.
What Claude Mythos 5.1 is
Claude Mythos 5.1 uses the same model as Fable 5.1. Anthropic says its safeguards are more permissive for vetted work in cybersecurity and the life sciences.
Access is not self-serve. Anthropic says Mythos 5.1 is available through trusted-access programs, including:
- the Cyber Verification Program (CVP)
- the Life Sciences Verification Program (LSVP), developed with the U.S. government
The company says Mythos 5.1 is currently available to a set of U.S. organizations, and that it is working with the U.S. government on wider access.
On Terminal-Bench 4.0, Anthropic reports 60.9% for Mythos 5.1.
Price and cache-read changes
Anthropic cut the price of cache reads — reuse of context the model has already processed — by 75%, to $0.25 per million tokens where usage is billed by token.
The company estimates that typical workloads cost about 25% less than on Fable 5, and that highly agentic coding tasks can cost up to about 45% less, because those jobs reuse cached context more often.
Safety considerations
Anthropic published a system card with the release. It says it tested chemical, biological, cyber, agentic, and alignment risks before launch.
On cybersecurity, the company says Fable 5.1 can now be used to identify software vulnerabilities. It still routes some dual-use tasks — including penetration testing, exploit generation, and binary-based vulnerability scanning — to Opus models.
On biology, Anthropic says Fable 5.1 is less likely to block basic biology and medical questions than Fable 5, while research-and-development life-science work remains restricted on the generally available model.
Enterprise Frontier Safeguards (EFS) is a separate system Anthropic says will store monitoring data in customer-controlled cloud infrastructure rather than on Anthropic’s systems. The company says EFS will roll out in phases starting later this fall. Until then, eligible customers can use Fable 5.1 with zero data retention.
Anthropic also says outputs from models released after 2 August 2026 include an invisible watermark under the EU AI Act Code of Practice on Transparency of AI-Generated Content.
Why the announcement matters
For developers, Fable 5.1 is another frontier coding model in a market already contested by Google, OpenAI, and others. See related coverage of reports about a coding-focused Gemini update.
The cheaper cache-read price is aimed at long-running agents that keep a large context in memory. That is a cost issue for teams that already run coding agents for hours, not only a benchmark change.
Mythos 5.1 matters for a smaller group: organizations that need the same model with fewer blocks on defensive security or professional life-science work, and that can pass Anthropic’s verification programs.
What it could mean for AI-powered coding and research
Anthropic argues these models show an early path for AI to help with scientific work. Its announcement describes company-run experiments in protein-binder design, a Venus elevation map built from NASA Magellan radar data, and GPU-kernel speedups on open-source biology models.
Those examples are Anthropic’s own research demonstrations. They are not independent clinical or mission results.
If the published coding and research scores hold up in everyday use, teams that already rely on Claude for software and analysis will have a newer, more expensive-tier option with lower cache-read costs. Broader scientific use of Mythos 5.1 still depends on who Anthropic admits to its access programs.
Key Takeaways
- Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on 1 September 2026.
- The company says they are the same model with different safeguards.
- Fable 5.1 is generally available for coding, knowledge work, and research, including via
claude-fable-5-1. - Mythos 5.1 is limited to trusted-access programs for cybersecurity and life sciences.
- Cache reads are 75% cheaper; Anthropic estimates typical savings of about 25%, and up to about 45% on highly agentic work.
- Fable 5.1 can be used to find software vulnerabilities, but some offensive or dual-use cyber tasks stay restricted.
