Anthropic’s Secret Model is Even Better Than Mythos 5 But You Won’t Get It

Anthropic has made public that it is already using an unreleased AI system called Model 2.

Key points

• Model 2 scored 62.8%, compared with 50.3% for Claude Mythos 5 and 54.8% for Mythos Preview.
• New DeepSeek V4 Pro is Almost as Good as Claude Fable 5 and Super Cheap.
• Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.
• The previously undisclosed model appeared in Anthropic's August 2026 Risk Report.
• The clearest performance difference appears in CoBench v2, Anthropic's internal benchmark based on 449 real research and engineering challenges previously solved by its employees.

Anthropic has made public that it is already using an unreleased AI system called Model 2. It the business describes as somewhat more capable than Claude Mythos 5.

In practice, the previously undisclosed model appeared in Anthropic's August 2026 Risk Report. As of the report's July 15 coverage date, Model 2 was one of three frontier or near-frontier models that had not been publicly published. Claude Can Now Send Gmail Messages Without Your Approval. Model 2 Beats Mythos 5 on Anthropic's Internal Test.

Meanwhile, the clearest performance difference appears in CoBench v2, Anthropic's internal benchmark based on 449 real research and engineering challenges previously solved by its employees.

Model 2 scored 62.8%, compared with 50.3% for Claude Mythos 5 and 54.8% for Mythos Preview. While Opus 4.6 reached 15.6%, claude Opus 4.7 scored 27.4%.

Anthropic estimates that a system capable of fully replacing its technical research staff would need to score at least 85% on CoBench, meaning Model 2 remains well below that threshold.

Notably, the firm describes Model 2 as a "noticeable improvement" over Mythos 5 for numerous internal tasks, but notes the jump is smaller than the earlier improvement from Claude Opus 4.6 to Mythos Preview.

New DeepSeek V4 Pro is Almost as Good as Claude Fable 5 and Super Cheap. Anthropic Is Already Using Model 2 Internally.

Model 2 is not merely an experimental benchmark model. Anthropic notes it and Mythos 5 are heavily employed internally for coding, data generation, research, engineering and other agentic workloads, including persistent AI-agent deployments.

Nevertheless, Anthropic states it at present has no aims to release Model 2 externally.

Meanwhile, the firm has not completed its usual full suite of predeployment assessments, meaning it has lower confidence in its understanding of the model's capabilities than it does for fully evaluated models.

Anthropic additionally stated its internal deployment review discovered no new or more concerning forms of misalignment compared with those already observed in Mythos 5.

There is at present no indication that Model 2 will eventually launch as Claude Mythos 6 or under any other commercial name.

Obtain the latest tech news, telecom insights, and product launches wherever you prefer.

Add as a preferredSource on Google Follow on Google News Join WhatsApp.

Technology and Automotive Specialist covering the latest cars, smartphones, AI breakthroughs, and.

Taken together, the developments around anthropic's Secret Model is Even Better Than Mythos 5 But You Won't point to a situation that is still moving, and the coming days should bring more clarity.