Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Anthropic The most capable AI model It’s still in the hands of a few organizations and cybersecurity professionals, but the most powerful cloud model you can use right now is getting an upgrade.
Claude Opus 4.8, released on Thursday, is a “modest but tangible improvement” over Opus 4.7, Anthropy said in a blog post.
But the company also said it is making significant progress in producing a version of the Claude Mythos Preview model that it wants to release to the public. For now, it has restricted access to Mythos to a group of partners as part of what it calls Glasswing ProjectExplaining that the model’s cybersecurity capabilities are advanced enough to justify giving cybersecurity experts and major technology companies some time to correct the flaws discovered by the model.
“Models with this level of capability require stronger cyber assurances before they can be released generally,” Anthropic said. “We are making rapid progress in developing these safeguards and expect to be able to deliver Mythos-class models to all of our customers in the coming weeks.”
Mythology for everyone?
Anthropic’s decision to withhold the Mythos Preview from the general public, at least for now, has been an interesting one. Was this a wise and forward-looking move to protect critical Internet infrastructure from potential flaws? Was it an easy way to generate marketing hype? Security researchers have found that this model is certainly capable of that Find exploits much more quickly From human hackers, even if not necessarily beyond human capabilities. The latest Mozilla version for Firefox is included Over 200 fixes have been identified by Mythos Preview.
Read more: The AI arms race is accelerating with new models from OpenAI, DeepSeek, and Anthropic
But the fact that Mythos will soon be available to anyone, even with significant cybersecurity guardrails, means we’ll finally see if the model lives up to the hype, with all the risks that might entail.
Darren Williams, founder and CEO of cybersecurity company Black fogtold CNET in an email that big model releases are often tense moments.
“On the one hand, Anthropic’s decision to release prisoners, and to hold off until safeguards are developed, shows the right instincts,” he said. “But the more capable the model, the greater the risk if those safeguards fail or if the model is eventually abused. The period between the launch of a robust model and the widespread adoption of defenses against it is always a moment of weakness.”
However, Mythos will be much more expensive to run than other AI models, and this may limit its usefulness to hackers. Jake WilliamsMythos was 30 times more expensive in testing than the previous Opus model, said a cybersecurity researcher and faculty member at IANS Research.
“This is beyond the reach of many, including any commodity threat actors,” Williams told CNET in an email. “Nation-state actors already had better technology to find vulnerabilities. This is only a game-changer for a small percentage of threat actors.”
What’s new in Cloud Opus 4.8
As for the Opus 4.8, Anthropic said it’s an improvement across benchmarks compared to the Opus 4.7. The company said tests found that Opus 4.8 was less likely to make unsupported claims and more likely to indicate uncertainty.
Some new features are also coming to Anthropic’s AI products, including the ability to control how much “effort” the model will use to respond to the prompt Claude.i And in Claude Cowork. A higher voltage will likely produce better results as the model spends more time responding, but it will exceed the limits of use faster. A lower setting will respond faster and reach rate limits slower.
