Anthropic's Claude Fable 5.1 Blocks Copycats With New Restrictions

Anthropic launched Claude Fable 5.1 on Tuesday, featuring a restriction on its reasoning process. Fresh developer accounts are barred from editing a conversation's history while preserving the model's saved thought process.

Anthropic describes Fable 5.1 and its limited version Mythos 5.1 as the most powerful models for coding and knowledge tasks. Both are now available on Anthropic's own platform, as well as Amazon Bedrock, Google Cloud, and Microsoft Foundry.

We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.

They're the world’s most advanced models for coding and knowledge work.

— Claude (@claudeai) September 1, 2026

What Claude Fable 5.1 costs and scores

The listed price remains unchanged. Fable 5.1 charges $10 per million input tokens and $50 per million output.

Cost reductions appear in other areas. Cache reads decrease 75% to $0.25 per million. Anthropic estimates typical workloads become 25% less expensive, with complex agent tasks dropping as much as 45%.

The biggest improvement on Anthropic's internal benchmarks is in agentic science tasks. Fable 5.1 scored 52.6% on Terminal-Bench-Science 0.1, more than twice Fable 5's 24.7%.

The coding advantage is slimmer. Fable 5.1 achieved 55.8% on Terminal-Bench 4.0, beating Opus 5's 52.3% and OpenAI's GPT-5.6 Sol at 37.3%.

On AutomationBench, business workflow scores almost doubled to 31.4%. The knowledge cutoff is now June 2026, five months more recent than Fable 5.

Why the copying protection matters

A subtler update is aimed at competitors. Anthropic made its intention clear in the launch announcement.

“It is no longer possible for new API accounts to manually edit Claude’s prior context in a multi-turn conversation while preserving the transcript of Claude’s prior thinking.”

That editing technique is a common method for distillation, where a less expensive model is trained on the outputs of a more advanced one.

Anthropic quantified the problem in February, tying over 16 million Claude interactions to distillation efforts carried out via about 24,000 fraudulent accounts.

The company identified three Chinese research labs: DeepSeek, Moonshot AI, and MiniMax. Moonshot was responsible for 3.4 million exchanges alone.

In July, White House science adviser Michael Kratsios escalated the issue, alleging that Moonshot copied Anthropic's flagship model to develop its Kimi K3 system. Moonshot has not responded publicly.

The new rule applies only to accounts created on or after August 31. Older accounts remain unaffected, and users of Claude Code, Cowork, and Claude.ai will see no change.

Anthropic states that all accounts will be subject to this check on future models. Fable 5.1 will remain available at least until September 1, 2027.