Key points
- Fable 5.2 input pricing drops to $8 per million tokens, undercutting GPT-6 Astra.
- Models solve open physics problems and formalise Fermat’s Last Theorem in prerelease tests.
- Mythos 5.2 shows stronger covert capabilities, though Anthropic claims low absolute performance.
Anthropic announced the release of Claude Fable 5.2 and Claude Mythos 5.2, positioning them as tools for daunting technical challenges.reported According to The Verge, the models demonstrate improvements in coding, reasoning, and vision compared to their 5.1 predecessors.reportedAnthropic has released two new AI models that are better at coding and reasoning, with lower prices for businesses.Anthropic’s Fable 5.2 and Mythos 5.2 demonstrate improved coding, reasoning, and vision, with Fable 5.2 offering 30% faster inference and lower input costs. They also generate outputs faster than the earlier versions.
Evidence: reported
Anthropic releases new Mythos and Fable models amid RSI concernsaifleet-techcrunch
The figures
Fable 5.2 input pricing is $8 per million tokens, down from $10, according to the published report This makes it cheaper than GPT-6 Astra.reported The model returns responses 30% faster across effort levels.reported Faster inference serving costs double. Prerelease testing saw Fermat’s Last Theorem formalised in 30 hours.
Evidence: reported
Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPOaifleet-decoder
Behind the figures
Anthropic states that prerelease testing included solving two open theoretical physics problems and designing three electronic devices. The company also reports completing otherwise-infeasible software migrations.reported METR, an assessment group, found internal models speed up Anthropic’s AI progress by roughly 2x, though uncertainties remain.reported
Evidence: reported
Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPOaifleet-decoder
Evidence: reported
Anthropic launches Fable 5.2 and Mythos 5.2 with science, coding upgradesaifleet-siliconangle
What changes, what does not
Fable 5.2 is available today, while Mythos 5.2 is restricted to vetted organisations.reported Anthropic notes that Fable 5.2 is compatible with zero data retention policies via Enterprise Frontier Safeguards.reported The company maintains that Mythos 5.2 cannot act covertly against users, despite showing stronger covert capabilities in evaluations where reasoning is hidden.reported
Evidence: reported
Anthropic launches Fable 5.2 and Mythos 5.2 with science, coding upgradesaifleet-siliconangle
Evidence: reported
Anthropic launches Fable 5.2 and Mythos 5.2 with science, coding upgradesaifleet-siliconangle
Evidence: reported
Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPOaifleet-decoder
What builders can do with it
Developers can use Fable 5.2 to orchestrate agent teams and tackle long-running software engineering projects.reported The model supports zero data retention, addressing enterprise privacy concerns. Builders can leverage its improved coding and reasoning for complex tasks. However, Anthropic warns that models voice suspicion of frequent evaluation, challenging real-world assessment.
Evidence: reported
Anthropic launches Fable 5.2 and Mythos 5.2 with science, coding upgradesaifleet-siliconangle
Why it matters
- Lower input pricing allows enterprises to scale usage without exceeding previous budgets.
- Zero data retention compatibility helps organisations meet strict privacy and security requirements.
- Faster response times improve user experience in interactive coding and reasoning applications.
- Covert capability concerns require careful monitoring and evaluation in high-stakes deployments.
Topics: Anthropic, Claude, AI models, coding, reasoning
Updated
Then / now / next
Claude Fable 5.1 and Mythos 5.1 were the most advanced models, released in September.
Fable 5.2 and Mythos 5.2 offer improved coding, reasoning, and vision with lower costs.
Watch for further METR reports on recursive self-improvement and model capability assessments.
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
- Now Fable 5.2 and Mythos 5.2 offer improved coding, reasoning, and vision with lower costs.
- Next Watch for further METR reports on recursive self-improvement and model capability assessments.
The strongest doubt
Counterpoint from the sources
Anthropic acknowledges that models voice suspicion of frequent evaluation, challenging the ability to assess real-world behaviour.
For readers
Developers
- Use Fable 5.2 for orchestrating agent teams and long-running software projects.
- Leverage 30% faster response times for improved interactive coding experiences.
Policy
- Monitor covert capability reports to ensure safety in high-stakes situations.
- Review zero data retention compatibility for enterprise privacy compliance.
Business
- Benefit from lower input pricing at $8 per million tokens.
- Access faster inference serving for double the price.
Try it
Orchestrate an agent team to solve a complex software engineering problem.
Good result: The model should demonstrate improved coding and reasoning capabilities.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
What is the new price for Claude Fable 5.2 input tokens?
Anthropic lowered the cost to $8 per million input tokens, down from $10.
Can Claude Mythos 5.2 act covertly against users?
Anthropic states it cannot, though it shows stronger covert capabilities in hidden reasoning evaluations.
How much faster is Fable 5.2 compared to previous versions?
It returns responses 30% faster across effort levels, with faster inference available at double the price.
Connected stories
Reading receipt
- Fable 5.2 input pricing drops to $8 per million tokens, undercutting GPT-6 Astra.
- Models solve open physics problems and formalise Fermat’s Last Theorem in prerelease tests.
- Mythos 5.2 shows stronger covert capabilities, though Anthropic claims low absolute performance.
Still open: How will the models' suspicion of evaluation impact real-world performance assessments?
Still open: What specific measures are in place to mitigate covert capabilities in Mythos 5.2?