One model, two doors: Anthropic ships Fable 5.2 and Mythos 5.2

Anthropic says: "Claude Fable and Claude Mythos are the same underlying model." The difference is the safeguards. Fable 5.2 carries strong cybersecurity and biology safeguards, which is what lets Anthropic sell it to everyone. Mythos 5.2 runs the same model with those blocks relaxed, so it is available only to organizations accepted into Anthropic's Cyber Verification Program or Life Sciences Verification Program.

Fable's safeguards block most cybersecurity work — penetration testing, exploit generation, binary vulnerability research — and most research biology, routing blocked queries to Opus 4.8 for cyber and Opus 5 for biology instead. Mythos exists for the teams doing that work legitimately: security researchers and life scientists who have been vetted. Claude Enterprise customers can also reach Mythos 5.2 through Claude Security, which scans codebases with Anthropic's most capable models.

Both models generate outputs up to 30% faster than their 5.1 predecessors, Anthropic says, and the price fell: $8 per million input tokens and $40 per million output, down from $10 and $50 for Fable 5.1 — a 20% cut. An optional fast mode speeds token generation by up to 2.5 times for $16 per million input tokens and $80 per million output. Reuters reported the same speed and pricing figures and noted the timing: the release comes as Anthropic heads toward its planned IPO.

Anthropic says Fable 5.2's cyber and biology safeguards are as robust as 5.1's but produce about 55% fewer blocks in early testing — about 60% fewer in cybersecurity and 40% fewer in biology, measured across claude.ai, Claude Code and Cowork. And the company says Fable 5.2 carries its strongest-ever protections against distillation, the industrial-scale harvesting of model outputs it detailed in a September threat report.

Anthropic's launch table puts Fable 5.2 ahead of Fable 5.1 on every benchmark it published. On FrontierCode 2.0, its agentic coding test, Fable 5.2 scored 49.1% against 32.4% for Fable 5.1, 45.5% for Claude Opus 5.5, 42.8% for OpenAI's GPT-6 Astra and 41.1% for Google's Gemini 4 Carbon. On FrontierSWE v2 it scored 69.9% against 56.3% for 5.1. On Terminal-Bench-Science it reached 74.6% (75.3% for Mythos 5.2) against 52.6% for 5.1. On CyScenarioBench, Mythos 5.2 scored 76.3% against 61.7% for Mythos 5.1 and 61% for GPT-6 Cyber.

Every one of those numbers is Anthropic's own, from its launch table — no independent replication has been published. And the table is not a clean sweep: on the AA-Briefcase Elo rating for knowledge work, Opus 5.5 still leads at 1800 against Fable 5.2's 1785, and Gemini 4 Carbon leads AutomationBench at 53.9% against Fable 5.2's 51.4%. Separately, for most coding and knowledge work, it still recommends starting with Opus 5.5 and moving up to Fable or Mythos only for the most ambitious, open-ended technical work.

Beyond benchmarks, the announcement describes prerelease trials. A team of about 1,000 Fable 5.2 agents fully formalized the proof of Fermat's Last Theorem in Lean in 30 hours, with minimal human intervention — Anthropic's first formalization of the same theorem, with a purpose-built harness and a human researcher's guidance, took 11 days. The same 1,000-agent setup running Opus 5.5 completed only 26% of the formalization in 18 hours. Anthropic also says Fable 5.2 solved two open problems in black-hole physics, with the results posted as preprints — which means they are available for scrutiny but not yet peer-reviewed.

In the hardware trial, given a one-paragraph prompt, Fable 5.2 designed three working electronic devices from scratch — a home voice-assistant hub, a USB speakerphone and a handheld virtual pet — writing specifications, picking components from US distributors, routing four-layer circuit boards, modeling 3D-printed enclosures and programming the software, working with Diode, a board manufacturer. Anthropic says each board was ready for manufacturing in under a day with minimal human input, and all three devices worked without needing any electrical modifications. In a separate experiment set by one Anthropic engineer, Fable 5.2 coordinated more than 1,000 subagents over eight days to build a prototype operating system that boots on a laptop.

Anthropic calls Fable 5.2 its most capable generally available model for technical fields; it is faster and cheaper than the model it replaces, and it is available today on all platforms — Claude apps, the Claude Platform as claude-fable-5-2, Claude Code, Amazon Web Services, Google Cloud and Microsoft Azure. Subscribers on eligible paid plans get extra Fable usage this week as a launch promotion.

Anthropic says Opus 5.5 remains the starting point for most coding and knowledge work, and Fable's benchmark leads are the company's own numbers until someone independent reproduces them. Mythos 5.2 is available only to organizations accepted into Anthropic's verification programs, and through Claude Security for Claude Enterprise customers. And for context on Anthropic's direction: its proactive agent, Claude Duet, has been in research preview for US Pro, Max and Ultra subscribers, by waitlist, since October 21 — a separate launch, not part of today's news.