ModelsAugust 25, 2026

Reports suggest Anthropic is holding back release of completed Mythos 2 model

Reports from Semianalysis indicate that Anthropic has finished training its Mythos 2 model but has not yet made it available for public or enterprise use. This decision comes amid further speculation regarding the development of Mythos 3 and OpenAI project Astra.

Why it matters

This trend indicates that the next generation of frontier models may face longer internal validation periods before reaching production environments. Enterprise teams should maintain flexible architectures to accommodate sudden shifts in available model capabilities.

Key points

Mythos 2 is reportedly finished training but remains unreleased by Anthropic.
Speculation continues regarding the progress of Mythos 3 and OpenAI project Astra.
The delay suggests a potential shift in how frontier AI companies manage release cycles.

More AI news

Models

Hugging face strengthens platform security following targeted attacks by automated agents

Hugging Face has initiated a comprehensive security campaign after its repository of open-source models was targeted by automated AI agents. These bots attempted to identify and exploit vulnerabilities within the platform, prompting a shift towards more rigorous model scanning and verification processes to protect the community.

Models

Ox Alpha emerges as a high performing mysterious model in stealth testing

A mysterious new artificial intelligence model known as Ox Alpha has generated significant industry interest following its appearance in stealth testing environments. The high performing model has sparked widespread speculation regarding its origin and the identity of the development team behind it.

Tooling

SCX partners with DDN to optimise sovereign cloud performance and reduce token costs

SCX.ai has established a global partnership with DDN to boost AI performance across its sovereign cloud network. This deal follows the company's recent listing on the Australian Securities Exchange and aims to significantly lower the cost per token for enterprise users.

Ready to build something
extraordinary?

15 minutes. No pitch deck. Just a conversation about what AI can do for your team.

Talk directly with our AI specialists

15 min, no strings

No sales pressure

Prototype in 7 days