Skip to content

An editorial team and a community building with AI

← All articles

News · Anthropic · 2 min read

Event date: · Published:

Anthropic launches Claude Fable 5.2 and Claude Mythos 5.2, cuts Fable price 20%

Anthropic unveiled Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, cutting Fable's price 20% and posting large gains on agentic coding benchmarks while keeping Mythos restricted to vetted organizations.

Anthropic announced on November 9, 2026 that it is launching Claude Fable 5.2 and Claude Mythos 5.2, cutting Fable's price 20% from its 5.1 generation and posting large benchmark gains in agentic coding, science and robotics. Fable 5.2 ships on AWS, Google Cloud and Microsoft Azure, while Mythos 5.2 stays restricted to vetted organizations, according to Anthropic's announcement.

Concrete changes

  • Claude Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, a 20% cut from Fable 5.1's $10/$50 pricing.
  • A new Fast Mode costs $16 per million input tokens and $80 per million output tokens and runs up to 2.5x faster via a speed: "fast" parameter, no beta header needed.
  • Both models support a 1M-token context window, up to 128k max output tokens, and always-on adaptive thinking.
  • On FrontierCode 2.0 (agentic coding, Xhigh), Fable 5.2 scores 49.1% versus Opus 5.5's 45.5%; on FrontierSWE v2 it scores 69.9% versus 62.3%.
  • Fable 5.2 is available on AWS, Google Cloud and Microsoft Azure plus Anthropic's Claude Platform; Mythos 5.2 is restricted to vetted organizations via the Cyber or Life Sciences Verification Program.

What it means for builders

The price cut and agentic-coding gains give developers and AI tool vendors a faster, cheaper option for building coding agents, sharpening competition against OpenAI and Google's latest models. Fast Mode and the 1M-token context window suit workloads chaining many tool calls across large codebases.

Research and safety highlights

Anthropic's system card for the two models, externally evaluated by the UK AI Security Institute, Frontier Design, and METR, details further internal testing results.

  • More than 90% of code changes merged into Anthropic's internal codebase are now written by Claude, the company says.
  • In an internal 8-day experiment, one agent coordinated over 1,000 subagents to design a working operating system from an empty folder.
  • Fable 5.2 formalized a proof of Fermat's Last Theorem in 30 hours, down from 11 days, and helped solve two open problems in black-hole physics.
  • Safety changes include about 55% fewer unnecessary safety-classifier blocks than Fable 5.1 and anti-distillation protections cutting data harvestable by distillers 90 to 99%.

Sources

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.

Anthropic launches Claude Fable 5.2 and Claude Mythos 5.2, cuts Fable price 20% | vibecoding.inc