Actually, Anthropic has an AI model called 'Model 2' that is more powerful than the top-of-the-line Mythos, but there are no plans to release it.

Anthropic has released a risk report on AI, and it has become a hot topic because it mentions a mysterious model that performs better than 'Claude Mythos 5,' the highest-performing model currently available to the public.

Redacted Risk Report August 2026 - Redacted Risk Report August 2026 .pdf
(PDF file)

As part of our Responsible Scaling Policy, we publish regular Risk Reports. These share detailed information on the risks of our systems and how prepared we are to address them.

Our second Risk Report is now available:

https://t.co/NgWnDmXZD3

— Anthropic (@AnthropicAI) August 14, 2026

A report published on August 15, 2026, shares the results of several safety reviews conducted on the Claude Mythos 5 and Model 2.

According to Anthropic, Model 2 is a model that 'has slightly higher capabilities than Claude Mythos 5' and is widely used within Anthropic alongside Claude Mythos 5. Claude Mythos 5 has been released to the public with additional security measures and is called 'Claude Fable 5,' but Anthropic states that there are no plans to release 'Model 2' to the public.

Model 2 has been evaluated by Anthropic employees and is perceived as 'clearly improved over Claude Mythos 5 in many tasks.' However, it does not show the same dramatic improvement as when switching from Claude Opus 4.6 to the original Claude Mythos 5 model, 'Claude Mythos Preview,' and is only considered 'slightly more capable overall.'

The results of Anthropic's internal benchmark, 'CoBench v2,' are as follows. Model 2 performed slightly better.

As a concrete example, in a test where the AI performed a secret task without being noticed by an observer, the success rate of Claude Mythos Preview was 6-10% with augmented thinking enabled, and the performance of Model 2 was slightly higher than Claude Mythos 5 but significantly lower than Claude Mythos Preview.

The report also mentions a model named 'Model 1.' One user pointed out that 'Model 2 is the successor to Model 1, and these training models are Claude Mythos Preview, while Claude Mythos 5 and Claude Fable 5 are smaller models created by distilling Model 2.'

I have no doubt in my mind anymore that Mythos Preview is their massive ~10T teacher model

I think Model 1 and Model 2 are further iterations of that same model

Mythos 5 and Fable 5 are likely only smaller distilled models https://t.co/J3OcqlPK9s pic.twitter.com/yXIxoeme8F

— Lisan al Gaib (@scaling01) August 14, 2026