Key Notes
- Claude Fable 5.2 and Mythos 5.2 use the same underlying model, with additional safeguards on Fable and direct Mythos access limited to vetted users and organizations.
- Anthropic says token generation is up to 30% faster than the 5.1 models, while standard API pricing falls to $8 per million input tokens and $40 per million output tokens.
- The company links the launch to two black-hole physics preprints, whose specific theoretical results require wider scientific scrutiny and do not constitute a complete theory of quantum gravity.
Anthropic has launched Claude Fable 5.2 and Claude Mythos 5.2, its most capable models yet, with improvements in coding, scientific reasoning and cybersecurity. The company announced the release on November 9, pairing faster output and lower prices with research results involving black-hole physics.
The two names represent different configurations of the same underlying model. Fable includes additional safeguards for sensitive cybersecurity and biology work, allowing broader access through paid Claude plans and developer services. Direct access to Mythos remains limited to vetted individuals and organizations through Anthropic’s verification programs.
Faster output at a lower price
Anthropic says the new models generate output tokens up to 30% faster than Fable 5.1 and Mythos 5.1. That measures generation speed; the time needed to finish a complex assignment also depends on reasoning, tool use and the number of steps involved. The release follows the Fable 5.1 update, which also emphasized performance and cost.
The updated Fable product page lists standard API pricing at $8 per million input tokens and $40 per million output tokens, down 20% from Fable 5.1’s $10 and $50. An optional fast mode costs $16 and $80 respectively, with token generation up to 2.5x faster. Fable 5.2 is available on Pro, Max, Ultra, Team and Enterprise plans, as well as through Anthropic’s API and supported cloud platforms.
In Anthropic’s FrontierCode 2.0 evaluation, Fable 5.2 reached 49.1% at its Xhigh effort setting, compared with 32.4% for Fable 5.1. At High effort, it scored 45.3%, roughly matching Opus 5.5’s best result of 45.5% at about half the cost per task. These are company-reported results under particular testing conditions, rather than a universal ranking for every coding assignment.
What Claude’s black-hole research actually found
The scientific work concerns holographic duality, a framework connecting certain theories of gravity to quantum theories without gravity. In the first preprint, researchers report a calculation addressing a question left open for more than two decades: whether a transition associated with black-hole formation remains abrupt in a weakly interacting quantum description. Their result says that it does.
A second preprint establishes the precise limit of a mathematical method used to constrain possible quantum-gravity theories in a simplified setting. Anthropic attributes both results to Fable 5.2. The papers’ provenance statements describe work by an internal Claude research model with checks by a physicist; the second also says Opus 5.5 independently obtained the main result.
Both papers are preprints, and Hexagon labels their text primarily AI-generated with human understanding of some parts. They address specific theoretical questions, rather than offering a complete theory of quantum gravity. Their significance will depend on scrutiny of the arguments and calculations by the wider research community.
Stronger cyber capabilities, with restricted access
Anthropic’s accompanying system card reports that Mythos 5.2 matched or exceeded its predecessors across the cybersecurity evaluations disclosed. It also documents limitations: Fable’s fallback to an older model in some prompt-injection scenarios increased attack success, even though Fable itself was more resistant when it answered directly. The release does not establish that autonomous agents can be trusted without oversight.
The company also constrained certain abilities that could directly accelerate frontier AI development during training. Its assessment is that Mythos 5.2 has not crossed its threshold for substantially accelerating its own AI research, while acknowledging uncertainty. That distinction matters after Anthropic’s recent self-improvement warning: better coding and scientific results do not, by themselves, demonstrate an uncontrolled cycle of AI building more capable AI.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
