we_are_coded.by CODE · The world, decoded
БГ
Anthropic

Fable 5.1 and Mythos 5.1 are the same model, and the only difference between them is the safeguards

Anthropic · event date: 1 September 2026Frontier

On 1 September Anthropic released Claude Fable 5.1 for everyone and Mythos 5.1 only for vetted cybersecurity and life sciences teams. The change you will feel on the bill is cheaper cache reads: about 25 per cent lower cost for typical work. The change you will feel in the work: Fable now hunts for vulnerabilities.

In short
  • Cache reads get 75 per cent cheaper, down to $0.25 per million tokens; input and output stay at $10 and $50.
  • The new cyber safeguards mean about 60 per cent fewer interventions per session in Claude Code. Fable 5.1 finds vulnerabilities but does not write exploits.
  • Enterprise Frontier Safeguards keeps data in the customer's cloud; it arrives in phases from the autumn, and until then eligible customers get zero data retention.
Checked on1 October 2026Responsible editorTsvetelin IvanovHow we workMethod · Corrections

A crash that happens once in a million runs and that nobody on the team had explained in four to five years. According to the investment firm Millennium, Fable 5.1 is the first model to find it: it disassembled a third-party library, matched it against the core dump and got to the bug.

Stories like this are the most interesting part of model announcements and the hardest to check. This one comes from a customer quoted by Anthropic. I include it because it is concrete, not because it is proof.

The facts: on 1 September 2026 Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1, the same model with different levels of safeguards. Fable 5.1 is generally available on all platforms, including AWS, Google Cloud and Microsoft Azure; Mythos 5.1 is offered only through trusted access programmes for cybersecurity and life sciences, and for now only to US organisations. Per the company, Fable 5.1 costs about 25 per cent less than Fable 5 for typical workloads and up to about 45 per cent less for highly agentic work, because cache reads get 75 per cent cheaper, down to $0.25 per million tokens; input and output pricing stays at $10 and $50 per million tokens. The new cyber safeguards mean around 60 per cent fewer interventions per session on average in Claude Code, and Fable 5.1 can now be used to discover vulnerabilities, but not to develop exploits; penetration testing and exploit generation are redirected to Opus models. Anthropic announced Enterprise Frontier Safeguards, a system in which data is stored in cloud infrastructure controlled entirely by the customer; it arrives in phases from the autumn, and until then eligible customers can use Fable 5.1 with zero data retention. Per Anthropic, Fable 5.1 scores 55.8 per cent on Terminal-Bench 4.0 against 42.0 for Fable 5. New API accounts can no longer edit Claude's prior context while preserving the transcript of its thinking - a measure against distillation.

The benchmark table is long and it is theirs. I will not copy it.

The bill is the news

In highly agentic work cache reads make up most of the bill: the agent goes back to context already loaded. So a 75 per cent cut exactly there turns into about 25 per cent for ordinary work and up to 45 for agentic work. Input and output are untouched. What has been touched is where agents spend.

They did not cut the price of the model. They cut the price of its habit of rereading.

The second change is for security people. Until now Fable stopped many of your questions if they were about vulnerabilities. Now it will look for holes in your code. It will not write you an exploit, and the penetration test goes to Opus.

Paying for Fable by the token? Switch to 5.1 and the bill drops by itself. Dropped it because of security refusals? Try again with a real task. And if you have your own integrations that edit Claude's prior context, read their note on distillation before you open a new account.

The visual is generated code art. No third-party images.
Follow usFacebookLinkedIn
Official primary sources
→Anthropic - Introducing Claude Fable 5.1 and Claude Mythos 5.1, 01.09.2026→Anthropic - Developing Enterprise Frontier Safeguards with our customers, 01.09.2026
Original: https://wearecoded.com/en/articles/claude-fable-5-1-i-mythos-5-1.html
ShareFacebookXLinkedInTelegramWhatsApp
← Back to all news