Anthropic Plans Broad Mythos Rollout, Launches Opus 4.8 Amid $900B Valuation

A Turning Point for Anthropic

Anthropic announced on Thursday that it plans to make its powerful Mythos-class AI models available to all customers in the coming weeks, a significant shift for a model previously restricted to a small group of cybersecurity organisations. At the same time, the company launched Claude Opus 4.8, described as a modest but tangible upgrade to its most advanced publicly available model.

Opus 4.8: A Behavioural Upgrade

Opus 4.8 is available at the same pricing as its predecessor — $5 per million input tokens and $25 per million output tokens. Anthropic frames it primarily as a safety and reliability improvement: the model is roughly four times less likely than Opus 4.7 to let flaws in its own code pass without flagging them, and its misaligned-behaviour rates now approach those of Claude Mythos Preview, which Anthropic calls its best-behaved model.

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

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 organisations — including Google, Cisco, CrowdStrike, Palo Alto Networks, and Microsoft — used the model to identify and patch vulnerabilities in widely deployed software. 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 ability to close the gap between vulnerability discovery and exploitation is exactly why Anthropic initially withheld it from general release, choosing instead to develop safeguards against misuse.

A Shifting Timeline

As recently as late May, Anthropic had suggested Mythos-level models would be broadly accessible within the next 6 to 12 months, acknowledging that no organisation had yet built safeguards robust enough to prevent abuse. Thursday's announcement of a weeks-long timeline suggests the company believes it has made faster progress on those protections than expected.

A Record Valuation

The news arrives on the same day Anthropic announced it had raised $65 billion in new funding at a $900 billion valuation, surpassing OpenAI to become the most highly valued AI startup in the world. The combination of a record raise and a major capability rollout cements the company's position at the centre of the AI race.