FreshSynapse

AI trends & daily updates, in plain English

AI Models

Anthropic Launches Claude Fable 5.2 and Mythos 5.2 at Lower Prices — Same Model, Two Safety Levels

By FreshSynapse Editorial · · 4 min read

Close-up of a computer monitor showing colourful HTML code with pink, green and yellow syntax highlighting on a dark screen

Photo: Hippopx (public domain). Illustrative image.

What happened. Anthropic released two new models on Monday 9 November 2026: Claude Fable 5.2 and Claude Mythos 5.2. The company describes them as its most capable models for difficult, open-ended technical work — coding, multi-step reasoning, robotics control and scientific research — and says they are a step above Claude Opus 5.5 in almost every capability it evaluated. The launch comes the week Anthropic is expected to start the investor roadshow for its planned stock-market listing, and “fable 5.2” was trending on Google on Tuesday morning, up 700% in searches over the past day.

One model, two levels of access. Fable 5.2 and Mythos 5.2 share the same underlying model, the same 1-million-token context window, the same 128,000-token output limit and the same price. What differs is the safeguards. Fable blocks most high-risk cybersecurity work — penetration testing, exploit generation and binary vulnerability analysis — and most research-biology requests, which is what lets Anthropic offer it to everyone. Mythos drops those blocks and is available only to organisations accepted into Anthropic’s Cyber Verification Program or Life Sciences Verification Program. Enterprise customers can also reach Mythos through Claude Security, which scans codebases for vulnerabilities.

Cheaper and faster, according to Anthropic. Standard pricing is $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for the 5.1 versions — a 20% cut that also undercuts OpenAI’s GPT-6 Astra, listed at $10 and $50. An optional fast mode costs $16 and $80 and is said to generate tokens up to 2.5 times faster. Anthropic says both models generate output up to 30% faster than their 5.1 predecessors, and that early testers found they communicate more clearly. Both figures are Anthropic’s own claims: no independent measurement of the speed improvement has been published.

The benchmark picture. On Anthropic’s FrontierCode 2.0 coding benchmark, Fable 5.2 scored 49.1%, against 32.4% for Fable 5.1 and 45.5% for Opus 5.5. On the company’s internal CVE Rediscovery cybersecurity evaluation, Mythos 5.2 found 89.6% of known vulnerabilities, up from 78.3% for Mythos 5.1. Anthropic also says the models produce about 55% fewer unnecessary refusals than the 5.1 versions — about 60% fewer on cybersecurity tasks and 40% fewer on biology. These are vendor-reported scores from Anthropic’s own evaluations, not independent results.

The eye-catching experiments — read as claims. Anthropic’s announcement leans on pre-release experiments run by its own teams: a team of about 1,000 Fable 5.2 agents built a prototype operating system that could boot on a laptop in eight days, and a similar team formalised the proof of Fermat’s Last Theorem in the Lean proof system in about 30 hours. The company also says the models solved two previously open problems in black-hole physics and designed three working electronic devices from scratch. None of these is an independent finding. Formalising Fermat’s Last Theorem means translating an existing proof into machine-checkable form — it is not a new proof of the theorem — and a prototype operating system built in a guided experiment is not a production system. The preprints behind the physics claims have not been through peer review.

Where it is available, and what Anthropic still recommends. Fable 5.2 is available now on paid Claude plans, in Claude Code and on the Claude Platform as claude-fable-5-2, and through Amazon Web Services, Google Cloud and Microsoft Azure. Cognition made it available in its Devin coding tools on launch day. Anthropic still recommends that most users start with Claude Opus 5.5 for everyday coding and knowledge work, and reach for Fable or Mythos only for the most ambitious, long-running technical projects. The models require 30-day data retention and are not available under zero data retention unless Anthropic expressly authorises it.

Why this launch matters beyond the models. CNBC framed the release as Anthropic’s first frontier launch since chief executive Dario Amodei’s September essay calling for the industry to slow the pace of capability gains — a tension the company has not resolved publicly. The timing is also commercial: the models land as Anthropic heads into its IPO roadshow, with the company needing to show both frontier capability and a path to revenue. A 20% price cut on its most powerful model, in the same week, is a statement about where Anthropic thinks the competition is heading.

Sources

  • Anthropic
  • Claude
  • Fable 5.2
  • Mythos 5.2
  • AI models
  • coding agents

Related articles