Anthropic leapfrogs OpenAI as the most valuable AI startup — and drops a new model

• Anthropic released Opus 4.8, its latest general-purpose model, and closed a $65 billion fundraise.
• The company's new $965 billion valuation brings it above OpenAI's as the rivals eye IPOs.
• Opus 4.8 is a "modest" improvement, Anthropic says, adding that Mythos is on the way.

Anthropic has arrived with the newest entrant in the AI race — and a new stack of cash.

On Thursday, the AI lab announced a new $65 billion fundraise, valuing it at a staggering $965 billion and making it the most valuable AI startup — surpassing its rival OpenAI.

The same day, Anthropic released Claude Opus 4.8, a new version of its general-purpose large language model. Anthropic called it a "modest but tangible improvement on its predecessor" and said it's better at coding, reasoning, and general knowledge work.

Anthropic is in the midst of a high-stakes fight for technological supremacy and market share with OpenAI, and this new model marks the latest in a series of salvos. Each lab is gearing up to go public, pushing to sell its coding tools to as many developers as possible, and angling to secure enough computing power to sustain growth.

In a post about the fundraise, Anthropic said the new money will help it "expand compute to meet growing demand for Claude," among other efforts.

Anthropic's latest valuation is a huge leap from its $380 billion valuation in February. OpenAI's last fundraise in March valued the company at $852 billion; its last model release, GPT-5.5, came in April.

What's new about Anthropic's Opus 4.8 model

Anthropic chose this model release to launch a new "effort control," which lets users toggle between settings that determine how much computing power a model devotes to an answer. Companies have begun to reckon with their AI spending, and this tool will help engineers use their Claude Code budgets more slowly, for example.

Opus 4.8's fast mode is three times cheaper than the previous versions on Anthropic's models, the company wrote, offering another way for customers to cut costs.

Opus 4.8 is not the much-awaited Mythos model that Anthropic held back due to cybersecurity concerns.

Anthropic had kicked off a worried frenzy in the cybersecurity community in April when it said that Mythos was far better than past models at finding vulnerabilities to hack. It says it has been working with trusted companies and organizations, particularly in cybersecurity, as it develops guardrails for Mythos.

"We're making swift progress on developing these safeguards and expect to be able to bring Mythos-class models to all our customers in the coming weeks," the Thursday announcement said.

Anthropic called the new Opus 4.8 model "weaker" than Mythos and said it's "substantially behind" the unreleased model on cyber capabilities. The company also wrote that while Mythos always admits it's an AI, Opus 4.8 could be tricked into saying it's human in about 3% of tests.

Have a tip? Contact this reporter via email at scouncil@businessinsider.com, or over text, Signal, Telegram, or WhatsApp at 415-757-8198. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely.