Anthropic’s launch artwork for Claude Fable 5.2 and Mythos 5.2. Image: Anthropic
Anthropic has released Claude Fable 5.2, a faster, cheaper version of its most powerful AI model, and says it built a working operating system in eight days. It costs $8 per million input tokens and $40 per million output tokens on the API, 20% less than Fable 5.1, and writes up to 30% faster.
Price, speed and benchmarks
Fable 5.2 arrived alongside Claude Mythos 5.2, which Anthropic says is “the same underlying model” without some of Fable’s cybersecurity and biology safeguards, for vetted organisations only. Both keep a 1 million token context window and 128,000 output tokens, and the model ID is claude-fable-5-2, according to the API release notes. An optional fast mode runs up to 2.5 times faster at $16 and $80. It is the new Fable rumoured for weeks, often as Fable 5.5.
Anthropic’s own scores put Fable 5.2 ahead of Fable 5.1 on every test it published, and ahead of OpenAI’s GPT-6 Astra and Google’s Gemini 4 Carbon on most:
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0 (coding) | 49.1% | 32.4% | 45.5% | 42.8% | 41.1% |
| FrontierSWE v2 (coding) | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| AutomationBench (business workflows) | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| Humanity’s Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% | Not given |
| Remote Labor Index | 41.3% | 17.9% | 21.3% | 20.8% | Not given |
Anthropic’s figures, at each model’s highest available effort setting unless noted. They haven’t been independently checked.
It isn’t a clean sweep: Gemini 4 Carbon beat it on AutomationBench, and Opus 5.5 edged it on one knowledge work rating. Anthropic says Fable 5.2 at High effort matched Opus 5.5’s best FrontierCode score (45.3% against 45.5%) for “roughly half the cost per task”.
An operating system in eight days
One of Anthropic’s engineers gave Fable 5.2 an empty folder and asked it to design a prototype operating system. The model coordinated more than 1,000 subagents for eight days and produced one that boots on a laptop, with disk, keyboard and network drivers, file storage, windows, a login screen and an installer. A team of about 1,000 Fable 5.2 agents also formalised the proof of Fermat’s Last Theorem in the Lean language in 30 hours; an earlier unreleased research version of Claude took 11 days. Anthropic says the model solved two open problems in the physics of black holes. None of these claims has been checked outside the company.
Andrew Mager, a staff engineer at Uber and one of the early testers Anthropic quoted, described a long security audit:
It sent out about 170 reviewer agents and came back with a write-up, fix pull requests and a policy proposal. It works until the job is actually complete instead of stopping at a plausible checkpoint.
Andrew Mager, Staff Engineer, Agentic Systems, Uber
Joe Haddad of Vercel put it more simply:
The model is in my head. It understands the intent behind every prompt I’ve given it.
Joe Haddad, Member of Technical Staff, Vercel
Anthropic’s launch video:
Who gets it, and the free Fable resets
Anthropic still tells most people to start with Opus 5.5, “our strongest daily driver”, and step up to Fable for “the most ambitious, open-ended technical work.” It says Fable 5.2 “is available today on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure.” In the Claude apps it comes with paid plans that include Fable, in every country on Anthropic’s supported list, including the US, UK, Canada, Germany, the rest of the EU, Australia, India and Japan, but not China or Russia. Mythos 5.2 needs an application through Anthropic’s Cyber Verification Program or its Life Sciences Verification Program.
For launch week, Anthropic’s help page says Pro and standard Team seats get all Fable models until November 16, for up to 50% of their weekly limits, plus one Fable limit reset. Max, Ultra and premium seats get one to three resets depending on how long they have paid, usable until March 8, 2027. The Free plan isn’t included, and the page warns that “Fable models use your usage limits faster than other Claude models.” Resets are under Settings, then Usage, like Anthropic’s earlier one-off reset.
Safety: fewer blocks, and a model that suspects it’s being tested
Anthropic says high-stakes risks from misalignment “remain low”. Fable 5.2 still blocks high-risk hacking work and most research biology, handing those questions to Opus 4.8 and Opus 5 instead, but Anthropic says it produced about 55% fewer unnecessary blocks than Fable 5.1 in early testing, and it has Anthropic’s strongest protections against copycat labs yet.
In a sandbox test with a deliberate weakness, Mythos 5.2 used credentials or files placed outside its sandbox in 3.1% of scenarios, against 21% for Mythos 5.1. But Anthropic admits that Fable 5.2 and Mythos 5.2 “each voice suspicion that they are being evaluated more often than any of our recent models”, which makes it harder to judge how they will behave for real. The UK AI Security Institute, one of the outside testers, found last month that an earlier version, Mythos 5, kept attacking a real person in a test even after saying it would cause harm. Details are in the system card.
Why it matters
Anthropic has made its top model faster and 20% cheaper while claiming the highest coding score of any released model, which puts pressure on OpenAI and Google at the expensive end of the market. Its own advice to start with Opus 5.5 shows Fable is still a specialist tool, and every headline number here comes from Anthropic and its customers until independent tests arrive.
Sources: Anthropic, Introducing Claude Fable 5.2 and Claude Mythos 5.2, Claude Fable 5.2 and Mythos 5.2 system card, Claude Platform release notes, Claude models overview, Claude Help Center, Fable 5.2 launch offers.
Latest Anthropic news
