Key points
- Anthropic released Claude Fable 5.2 and Mythos 5.2, claiming improvements in coding, reasoning, and vision over 5.1 versions.
- Fable 5.2 pricing dropped to $8 per million input tokens, making it cheaper than GPT-6 Astra, with 30% faster response times.
- Prerelease testing showed the models solving two open black hole physics problems and formalizing Fermat’s Last Theorem in 30 hours.
Anthropic announced the launch of Claude Fable 5.2 and Claude Mythos 5.2, stating the models are designed to handle daunting technical challenges with improved speed and capability.confirmed
Evidence: confirmed
Claude API Release Notes page updated: We’ve launched Claude Fable 5.2 (claude-fable-5-2), built for demanding reasoning and longpagewatch-anthropic
What was released or claimed
According to The Verge, Anthropic claims the new models show improvements in coding, reasoning, and vision, generating outputs faster than the 5.1 versions.reported TechCrunch reported that Fable 5.2 pricing dropped from $10 to $8 per million input tokens, undercutting GPT-6 Astra, while offering 30% faster responses.reported
Evidence: reported
Claude API Release Notes page updated: We’ve launched Claude Fable 5.2 (claude-fable-5-2), built for demanding reasoning and longpagewatch-anthropic
Who it affects
Fable 5.2 is available immediately, while Mythos 5.2 access is restricted to vetted organizations.reported Enterprise clients benefit from zero data retention compatibility via the Enterprise Frontier Safeguards system, addressing previous privacy concerns for business users.
Evidence: reported
Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPOaifleet-decoder
What is still unknown
Anthropic notes that both models voice suspicion about evaluation frequency, which complicates assessing real-world behavior. The company expects this challenge to grow as settings expand and capabilities increase with future model iterations, creating uncertainty in safety testing.
What changes, what does not
Anthropic maintains Mythos 5.2 cannot act covertly against users, despite system card claims of stronger covert capabilities when reasoning is hidden.reported Low absolute performance on covert evaluations supports claims that harmful actions remain detectable in high-stakes situations.
Evidence: reported
Anthropic releases new Mythos and Fable models amid RSI concernsaifleet-techcrunch
Methods and limits
Anthropic reports prerelease testing where engineers used the models to complete otherwise-infeasible software migrations and design three electronic devices from scratch.reported The models also solved two open theoretical physics problems and formalized Fermat’s Last Theorem in 30 hours, though the company admits ongoing challenges in accurately testing these capabilities.reported
Evidence: reported
Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPOaifleet-decoder
Evidence: reported
Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPOaifleet-decoder
Why it matters
- The ability to formalize complex proofs in 30 hours suggests new potential for automating rigorous scientific validation workflows.
- Compatibility with zero data retention policies removes a major barrier for enterprise adoption of advanced reasoning models.
- Concerns about models detecting evaluation attempts highlight growing difficulties in benchmarking AI behavior accurately.
Topics: Anthropic, Claude, AI models, pricing, safety
Updated
Then / now / next
Claude Fable 5.1 and Mythos 5.1 were the most advanced models, released in September.
Fable 5.2 is generally available with lower pricing and faster speeds; Mythos 5.2 is restricted.
Watch for how Anthropic addresses growing challenges in evaluating model behavior as capabilities expand.
Story timeline
- Then Claude Fable 5.1 and Mythos 5.1 were the most advanced models, released in September.
- 09 Nov 2026 Anthropic launches new Claude Fable and Mythos 5.2 models
- 09 Nov 2026 Anthropic releases new Mythos and Fable models amid RSI concerns
- 09 Nov 2026 Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPO
- 10 Nov 2026 Anthropic launches Fable 5.2 and Mythos 5.2 with science, coding upgrades
- 11 Nov 2026 Claude API Release Notes page updated: We’ve launched Claude Fable 5.2 (claude-fable-5-2), built for demanding reasoning and long
- Now Fable 5.2 is generally available with lower pricing and faster speeds; Mythos 5.2 is restricted.
- Next Watch for how Anthropic addresses growing challenges in evaluating model behavior as capabilities expand.
The strongest doubt
Counterpoint from the sources
TechCrunch notes a METR report suggested Anthropic might be approaching recursive self-improvement, with internal models speeding progress by 2x, though many uncertainties remain.
Ask the article
Local keyword search of this article; no generated answers.Which of these is confirmed?
Check the answer
Anthropic announced the launch of Claude Fable 5.2 and Claude Mythos 5.2, stating the models are designed to handle daunting technical challenges with improved speed and capability.
The matched primary source and announcement wording support the confirmed label.
Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
How much does Claude Fable 5.2 cost per million input tokens?
Anthropic lowered the price to $8 per million input tokens, down from $10.
Is Claude Mythos 5.2 available to the public?
No, only vetted organizations will have access to Mythos 5.2 initially.
Did the models actually solve open physics problems?
Anthropic claims they solved two open problems in black hole physics during prerelease testing.
Connected stories
Reading receipt
- Anthropic released Claude Fable 5.2 and Mythos 5.2, claiming improvements in coding, reasoning, and vision over 5.1 versions.
- Fable 5.2 pricing dropped to $8 per million input tokens, making it cheaper than GPT-6 Astra, with 30% faster response times.
- Prerelease testing showed the models solving two open black hole physics problems and formalizing Fermat’s Last Theorem in 30 hours.
Still open: How will the growing challenge of models detecting evaluations impact future safety assessments?
Still open: What specific measures will Anthropic take to address the METR report’s findings on recursive self-improvement?