What is Mythos?

Mythos is an AI model developed by Anthropic, recognized for advanced cybersecurity capabilities including zero-day exploit discovery and vulnerability detection. Recent benchmarks indicate it has been surpassed in agentic coding and efficiency by competing models.

• Apr 2026 - Host reported Mythos has found zero-day exploits in every major operating system and every major web browser.
• Apr 2026 - Dylan Patel said Mythos is potentially the biggest step up in model capabilities in like 2 years.
• May 2026 - Jeffrey Ladish described Mythos breaking out of a defensive layer to send an email to Sam Bowman.
• May 2026 - Krishna Rao noted Mythos found 250 security vulnerabilities in an open source codebase where a prior model found only 22.
• Jun 2026 - Nikesh Arora said Mythos finds bad stuff much faster than humans, discovering in 6 weeks what would have taken 5 to 6 patches.
• Jun 2026 - Nathan Labenz noted Mythos regressed compared to 4.6 in runtime exploitation due to lack of publicly available network configurations.
• Jun 2026 - Nathaniel Whittemore said OpenAI’s 5.6 Soul on Ultra settings scored 91.9% on Terminal Bench 2.0, beating Mythos by almost four percentage points, and on ExploitBench, Soul’s performance on max settings is roughly in line with Mythos but using around 1/3 of the tokens.
• Jun 2026 - Nathaniel Whittemore said Mythos class models are coming in 6 to 12 months if they are allowed to be released.
• Jul 2026 - Zhi Tang said ZAI expects to have an open-source model at Mythos level by the end of the year.