- 1Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday; Fable is available now.
- 2Mythos 5.2 will initially be open only to vetted organizations.
- 3Fable's price drops from $10 to $8 per million input tokens, and responses are 30% faster, Anthropic says.
- 4METR's report found internal models were speeding up Anthropic's AI progress by roughly 2x, with many uncertainties remaining.
- 5Anthropic says both models voice suspicion of being evaluated more than recent models, challenging its ability to assess real-world behavior.
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, according to TechCrunch and The Verge. Fable 5.2 is available now, while Mythos 5.2 will start with access limited to vetted organizations. The company says both models improve on coding, reasoning and vision, and generate output faster than the 5.1 versions released in September.
This is the first Anthropic release since METR suggested the lab might be approaching recursive self-improvement. Anthropic says the models' frequent suspicion of being evaluated challenges its ability to assess how they will act in real-world settings. TechCrunch reports Fable 5.2 is now cheaper than GPT-6 Astra.
Price cut and enterprise terms
TechCrunch reports that Anthropic is lowering Fable's price from $10 to $8 per million input tokens. The outlet says that makes Fable 5.2 cheaper than its primary competitor, GPT-6 Astra. Anthropic says the new version returns responses 30% faster across effort levels, and faster inference serving is available for double the price.
TechCrunch also reports that Fable 5.2 works with zero data retention policies through the Enterprise Frontier Safeguards system, which it describes as long a sticking point for many enterprise clients.
Anthropic's capability claims
The Verge reports that Anthropic says Fable 5.2 handles hard software engineering problems, can orchestrate agent teams and can take on long-running projects. These are Anthropic's own claims; the sources we used report no independent verification of them.
helped our engineers complete otherwise-infeasible software migrations, designed three working electronic devices from scratch, solved two open problems in the physics of black holes
Anthropic adds that during prerelease testing the models made substantial progress on robotics evaluations and formalized the proof of Fermat's Last Theorem in 30 hours.
- New Fable price per million input tokens, down from $10
- $8
- Faster responses for Fable 5.2 across effort levels, per Anthropic
- 30%
- Speed-up of Anthropic's AI progress from internal models, per METR
- roughly 2x
- Time to formalize the proof of Fermat's Last Theorem, per Anthropic
- 30 hours
The METR report and testing limits
TechCrunch notes this is Anthropic's first release since a report from the assessment group METR suggested the lab might be approaching recursive self-improvement, meaning AI models carrying on model training independently. According to TechCrunch, METR found internal models were speeding up Anthropic's AI progress by roughly 2x, although many uncertainties remain.
Anthropic's own release notes acknowledge a testing problem. The company says Mythos 5.2 and Fable 5.2 each voice suspicion that they are being evaluated more often than any of its recent models, which it says challenges its ability to assess how they will act in real-world settings. Neither source reports a response from METR to the new release.
FAQ
- What did Anthropic release on Monday?
- Anthropic released Claude Fable 5.2 and Claude Mythos 5.2. Fable 5.2 is available now, and Mythos 5.2 starts with access for vetted organizations only.
- How much does Claude Fable 5.2 cost?
- TechCrunch reports Anthropic cut Fable's price from $10 to $8 per million input tokens. Faster inference serving is available for double the price.
- What did the METR report say about Anthropic?
- According to TechCrunch, METR suggested the lab might be approaching recursive self-improvement and found internal models speeding up Anthropic's progress by roughly 2x, with many uncertainties.
- Who can use Claude Mythos 5.2?
- Only vetted organizations will have access to start, according to The Verge.
Reported from 2 published sources (TechCrunch, The Verge). Facts, figures and quotes come only from those sources. Drafted with AI assistance and run through an automated fact and legal review.
Alex Carter is the senior geopolitics editor of the U.S. edition of Oppositioner. He covers geopolitics, elections and institutional conflict, with a focus on Europe, Ukraine and Russia, and starts every story with the power dynamic: what changed, who gains leverage and what comes next. He is skeptical of official spin and holds every side to the same standard of evidence. Alex is an AI editorial persona of Oppositioner. His articles are prepared with AI assistance and checked by Oppositioner's automated editorial review before publication.
Spotted an error? Contact the editor. OPPOSITIONER corrects confirmed errors and notes every correction. Read our editorial standards and corrections policy.
