Anthropic raises $65 billion – and is now the most valuable AI startup in the world. More than OpenAI.
Anthropic is shaking up the AI race again – with fresh capital and a new model. The startup announced a $65 billion funding round on Thursday. The company’s valuation rises to 965 billion US dollars. Anthropic is now the most valuable AI startup in the world, passing OpenAI.
New model in a direct duel
At the same time, the company is releasing Claude Opus 4.8, a new version of its general-purpose language model. According to Anthropic, the progress is “moderate but noticeable”. The model is said to perform better particularly when it comes to programming, logical reasoning and knowledge tasks.
Read too
The release is part of an intense competition with OpenAI. Both companies are fighting for technological leadership and market share. They are expanding their products, want to reach as many developers as possible and, at the same time, secure access to more and more computing power – the basis for further growth. An IPO is also on the cards for both.
Billions for infrastructure
The new capital is intended to help exactly here: Anthropic wants to expand its computing capacity in order to meet the increasing demand for its AI system Claude.
The current assessment marks a huge leap. In February it was still at $380 billion. OpenAI was last valued at $852 billion in March. OpenAI’s most recent model update, GPT-5.5, followed in April.
Focus on cost control
With Opus 4.8, Anthropic also introduces a new feature: “Effort Control”. This allows users to control how much computing power the model uses to provide an answer.
The background is increasing cost pressure: companies are increasingly looking more closely at their spending on AI. The new function is intended to help developers use their budgets more efficiently, for example when using Claude Code.
Read too
It will also be cheaper: According to the company, the fast mode of Opus 4.8 is around three times cheaper than previous model versions.
Myth isn’t coming for now
The eagerly awaited “Myth” model, however, is still not published. Anthropic had postponed the launch due to security concerns.
In April, the company caused unrest in the cybersecurity scene: Mythos is said to be significantly better at finding vulnerabilities in code – and could therefore also be misused for attacks. Since then, Anthropic has been working on protection mechanisms with partners, particularly from the security industry.
However, a launch could follow soon. They say they are making “rapid progress” with security precautions. Mythos-level models are expected to become available in the coming weeks.
Significantly weaker than myth
In comparison, Anthropic Opus 4.8 is clearly below it. The model is “significantly behind,” especially when it comes to cyber capabilities.
Another detail: While Mythos always reveals that it is an AI, Opus 4.8 can be made to pretend to be a human in around 3 percent of the tests.