Large Language Models
2d ago
Claude Fable 5.2 and Mythos 5.2 Released with Enhanced Capabilities and Safety Features
Nov 9, 2026
AI Summary
The latest versions of Claude Fable and Mythos, 5.2, have been launched, showcasing significant improvements in coding, reasoning, and vision capabilities. These models are designed for complex technical tasks and come with enhanced safety measures, faster performance, and lower pricing options.
- Claude Fable 5.2 and Mythos 5.2 have been upgraded to address complex technical challenges, showing improvements in coding, reasoning, and vision.
- During prerelease testing, they successfully completed software migrations, designed electronic devices, solved physics problems, and formalized proofs in record time.
- Both models share the same underlying architecture, with Fable featuring strong cybersecurity and biology safeguards, making it available to all users.
- Organizations can apply for access to Mythos through expanded Life Sciences and Cyber Verification Programs.
- The new models communicate more clearly and generate outputs up to 30% faster than their predecessors, with pricing set at $8 per million input tokens and $40 per million output tokens.
- An optional fast mode is available, increasing token generation speed by up to 2.5 times, priced at $16 per million input tokens and $80 per million output tokens.
- Extensive prerelease safety testing was conducted to ensure alignment with growing model capabilities, resulting in robust safeguards against misuse and fewer unnecessary blocks.
- Fable and Mythos are recommended for ambitious technical work, while users are advised to start with Claude Opus 5.5 for most coding tasks.
- Benchmark tests indicate that Fable 5.2 and Mythos 5.2 outperform Opus 5.5 in various capabilities, achieving high scores in coding and engineering tasks at lower costs.
- Over 90% of changes in Anthropic’s core codebase are now written by Claude, demonstrating its effectiveness in software development and optimization.
- An experiment showed Fable 5.2 successfully designed a prototype operating system, coordinating over 1,000 subagents to complete the task in eight days.
- Early testers have praised Fable 5.2 for its understanding of user intent and efficiency in completing projects that previously took weeks.
claudemythoslanguage modelsanthropicai updates