Anthropic accelerates Mythos rollout after massive funding boost

Anthropic on Thursday launched Opus 4.8 and said it plans to broadly release its restricted Mythos-class AI models within weeks, per Reuters.whtc
Mythos, previously limited to about 50 cybersecurity partners under Project Glasswing, has found thousands of zero-day flaws across major operating systems and browsers.iansresearch
Anthropic also announced $65 billion in new funding at a $900 billion valuation, according to the New York Times, surpassing OpenAI as the top-valued AI startup.nytimes

Anthropic Plans Broad Mythos Rollout, Launches Claude Opus 4.8

Anthropic on Thursday said it plans to release its powerful Mythos-class AI models to all customers in the coming weeks, while simultaneously launching Claude Opus 4.8, an incremental upgrade to its most advanced publicly available model.techcrunch

The announcement marks a turning point for the San Francisco-based AI company, which had previously restricted Mythos to a small coalition of cybersecurity organizations under its Project Glasswing initiative due to the model’s extraordinary ability to autonomously discover and exploit software vulnerabilities.arcticwolf

Opus 4.8: A Behavioral Upgrade

Opus 4.8, available Thursday at the same pricing as its predecessor — $5 per million input tokens and $25 per million output tokens — is framed by Anthropic as “a modest but tangible improvement” over Opus 4.7. The model is roughly four times less likely than its predecessor to let flaws in code it produces pass without flagging them, and its misaligned-behavior rates now approach those of Claude Mythos Preview, which Anthropic describes as its “best-behaved” model.substack

Alongside the model, Anthropic launched Dynamic Workflows, a feature in research preview that allows Opus to manage complex tasks across hundreds of parallel subagents. The company said Claude Code with Opus 4.8 can now “carry out codebase-scale migrations across hundreds of thousands of lines of code from kickoff to merge.”techcrunch

Mythos and the Path to Broader Access

Claude Mythos Preview was first unveiled in April as part of Project Glasswing, a cybersecurity initiative in which roughly 50 organizations — including Google, Cisco, CrowdStrike, Palo Alto Networks, and Microsoft — have used the model to identify and patch vulnerabilities in widely deployed software.iansresearch

According to Anthropic, Mythos has surfaced thousands of high-severity vulnerabilities across every major operating system and web browser, including zero-day flaws that had gone undiscovered for up to 27 years. The model’s capacity to collapse the gap between vulnerability discovery and exploitation prompted Anthropic to withhold it from general availability while developing safeguards.reddit

A Shifting Timeline

As recently as late May, Anthropic had indicated that Mythos-level models would be “broadly accessible within the next 6-12 months” and acknowledged that no organization had yet developed safeguards robust enough to prevent misuse. Thursday’s announcement of a weeks-long timeline suggests the company believes it has made faster progress than initially expected on those protections.mashable

The news arrives on the same day Anthropic announced it had raised $65 billion in new funding at a $900 billion valuation, surpassing OpenAI as the most highly valued AI startup.nytimes