Anthropic's biggest day: Opus 4.8 ships, $965B valuation confirmed, Mythos coming in weeks
On May 28–29, 2026, Anthropic shipped Claude Opus 4.8 (4× more honest on code flaws), raised $65B at a $965B valuation to become the world's most valuable AI startup, and confirmed Mythos-class models will reach all customers in the coming weeks.
Anthropic had a consequential Thursday. On May 28–29, 2026, the company shipped Claude Opus 4.8, confirmed a timeline for its Mythos-class model, and announced a $65 billion funding round that pushes its valuation to $965 billion — past OpenAI and into position as the world's most valuable AI startup. Each item would be a standalone story on a slow news week; arriving together, they mark the sharpest single-day acceleration in Anthropic's five-year run.
Claude Opus 4.8: the "honest" flagship
Anthropic's stated headline for Opus 4.8 is honesty: the model is roughly four times less likely than its predecessor (Opus 4.7) to let code flaws pass without flagging them.1 Early testers say it is more willing to surface uncertainties and less prone to projecting false confidence — a persistent failure mode across AI coding tools.
Benchmark gains are across the board:
| Eval | Opus 4.7 | Opus 4.8 |
|---|---|---|
| Terminal-Bench 2.1 (agentic coding) | 64.3% | 69.2% |
| Multidisciplinary reasoning (with tools) | 54.7% | 57.9% |
| Agentic computer use | 82.8% | 83.4% |
| Knowledge work score | 1,753 | 1,890 |
Pricing is unchanged at $5 per million input tokens and $25 per million output tokens. The model is live in claude.ai, Claude Code, and the API. New features shipping alongside it include an effort control in claude.ai and Cowork (trade speed for depth), dynamic workflows in Claude Code that let the model plan and spawn hundreds of parallel subagents for codebase-scale migrations, and a revamped Messages API that accepts system-instruction updates mid-task without breaking the prompt cache.2
Enterprise adopters have been quick to report real-world results. Harvey (AI for legal work) says Opus 4.8 is the first model to break 10% overall on its Legal Agent Benchmark all-pass standard. Databricks reports 61% cheaper token cost than Opus 4.7 on its Genie agent. Cursor saw improvements at every effort level on its internal CursorBench. Cognition, the company behind the Devin coding agent, said the model fixed tool-calling and comment-verbosity issues that appeared in 4.7.1
Mythos is coming — in weeks
The bigger technical announcement may be what comes after Opus 4.8. Anthropic confirmed that Claude Mythos-class models will roll out to all customers in the coming weeks, following a period of restricted access.3
Mythos sits a full capability tier above Opus 4.7 and can autonomously discover zero-day vulnerabilities and write working exploits — which is why it spent April in a controlled hold. In the interim, Anthropic and roughly 50 partners (including Apple, Google, Microsoft, and AWS) used Mythos Preview through Project Glasswing to find more than 10,000 high- or critical-severity vulnerabilities across critical software infrastructure.1
Anthropic says its safety mitigations for Mythos are now strong enough to support general release. The delay itself triggered a small-scale geopolitical incident — several governments worried about exposure to financial and critical infrastructure vulnerabilities — and reportedly led to a direct call from Vice President JD Vance to AI company executives.4
Anthropic's own engineering blog describes the decision-making behind the hold: the model's "blast radius" in April was judged too high to ship without additional containment work.5
$65B Series H: $965B valuation, past OpenAI
On the same day, Anthropic announced its Series H: $65 billion raised at a $965 billion post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital.4 That puts Anthropic ahead of OpenAI's current private valuation (OpenAI closed a $122B round at $852B in March6) and within touching distance of a trillion.
The jump from $380 billion — where Anthropic stood after its February Series G — to $965 billion in roughly three months is driven by a revenue story: annualised run rate has grown from approximately $1 billion at end-2024 to an estimated $30 billion in 2026, pulled primarily by enterprise adoption of Claude.1
CFO Krishna Rao, announcing the round, pointed to Claude Code and Cowork as the products doing the heaviest lifting: "Claude is increasingly indispensable to our growing global community of customers, and we work tirelessly to make tools like Claude Code and Cowork more helpful, more powerful, and more adaptable to their needs."7
The valuation also carries IPO implications. Both OpenAI and SpaceX are expected to pursue public listings this year; Anthropic at $965B changes the comparative framing for any investor evaluating the sector.
What to watch next
Three things to track in the weeks ahead:
- Mythos general release date: Anthropic has said "coming weeks" but given no specific date. Any further delay — or any incident during the rollout — would be material.
- Pentagon lawsuit resolution: Anthropic is still fighting the Trump administration in federal court over the Pentagon's designation of Claude as a national-security risk. An appeals court heard arguments on May 19.8 Outcome unknown.
- Authors' copyright settlement: A federal judge is still reviewing the $1.5 billion Bartz v. Anthropic settlement. Authors who opted out have filed a separate suit requesting a jury trial.9
References
- 1Claude Opus 4.8 launch — The Next Web
- 2Claude Opus 4.8 release — Anthropic
- 3Anthropic confirms Mythos rollout — BleepingComputer
- 4Mythos geopolitical context — The Guardian
- 5How we contain Claude — Anthropic Engineering
- 6OpenAI $852B valuation — WSJ
- 7Anthropic Series H blog post — Anthropic
- 8Anthropic Pentagon lawsuit — The Hill
- 9Anthropic copyright settlement — Law.com
Add more perspectives or context around this Drop.