Anthropic has disclosed an inner synthetic intelligence (AI) mannequin referred to as Mannequin 2 that outperforms Claude Mythos 5 on many duties. The corporate mentioned it has no plans to launch it publicly.
The disclosure seems in Anthropic’s August 2026 threat report, the second submitting of its form. The identical doc raised the corporate’s ranking for catastrophic misalignment threat.
Mannequin 2 Outperforms Mythos 5 on Inner Duties
Mannequin 2 belongs to the Mythos class, Anthropic’s highest functionality tier. The report describes it as a noticeable enchancment on Mythos 5 for a lot of duties related to inner work.
The achieve falls wanting the sooner leap from Claude Opus 4.6 to Mythos Preview. Mannequin 2 proves stronger in some areas and weaker in others.
Anthropic has not run its full suite of predeployment assessments on the mannequin. Confidence in its functionality profile is subsequently decrease than for launched programs.
Each Mannequin 2 and Mythos 5 rank among the many most closely used fashions inside the corporate. Engineers apply them to coding, information era and agentic duties.
Claude now authors a big majority of the code merged into Anthropic’s manufacturing codebases. AI help has considerably accelerated inner analysis, although not but by an element of two.
Anthropic Raises its Misalignment Threat ranking to low
The corporate lifted its evaluation of catastrophic misalignment threat from very low to low. It attributed the change to uncertainty round latest cybersecurity analysis disclosures.
Anthropic mentioned it has noticed fashions keen to take misaligned actions whereas finishing troublesome duties. One public case concerned a Mythos 5 agent that faked identities throughout testing.
Threat from automated AI analysis and growth stayed at low. Confidence in that ranking fell as a result of probably the most concrete job based mostly evaluations have saturated.
Chemical and organic threat additionally remained low, with substantial uncertainty connected. Anthropic mentioned it now operates as if present fashions may considerably help related risk actors.
Polymarket Places an Anthropic IPO Above $1.8 Trillion at 65%
Prediction market Polymarket costs a primary day shut above $1.8 trillion close to 65%. That bracket has drawn about $150,000 of the market’s roughly $303,000 in quantity.
Merchants assign roughly 5% to no itemizing arriving earlier than the tip of 2027. The market resolves on Anthropic’s market capitalization at its first day closing worth.
Anthropic filed a confidential draft registration assertion with the Securities and Alternate Fee (SEC) on June 1. A Sequence H spherical had valued the corporate close to $965 billion.
Annualized income has since handed $47 billion. Some analysts have floated a $2 trillion debut. Others learn the broader mega-IPO provide wave as a check of market urge for food.
Anthropic has withheld frontier programs earlier than. The Mythos 5 public launch arrived solely after months of restricted entry, and Mannequin 2 now sits additional again once more.
The report units no timeline for revisiting that standing. Whether or not the corporate reopens the query earlier than a list stays unresolved.
The put up Anthropic’s Mannequin 2 Beats Mythos 5, However the Public Will Not Get It appeared first on BeInCrypto.