Anthropic's Claude Opus 5.5 lands the same day as GPT-6 Sol — cheaper, faster, and gated on cyber

On 22 September — the same day OpenAI released GPT-6 Sol and Luna — Anthropic released Claude Opus 5.5, the first model in its new 5.5 family. Both companies launched cheaper models on the same day, weeks after both of their CEOs publicly floated slowing the pace of frontier AI.

What's confirmed

The safeguards story

The more interesting detail is what Anthropic restricted. The company says Opus 5.5 matches its restricted-access Mythos 5.1 model in biology and cybersecurity capability, so it's applying safeguards similar to those on Fable 5.1. Reporting from the launch says most cybersecurity tasks are being re-routed to the older Opus 4.8, with access to cyber capabilities extended through a verification programme in the coming weeks.

That's the same pattern we covered in the cyber-AI arms race: the most capable models are shipping with their most dangerous capabilities deliberately gated.

Anthropic also says two outside safety organisations evaluated the model before release, and that it scored better than any recent Claude model on its internal audit of misaligned behaviour across nearly 2,000 scenarios.

Where it sits in the market

Opus 5.5 lands between OpenAI's premium Astra and its mid-tier Sol on price. The two labs are now competing across the whole price range rather than just at the top.

The slowdown question

Anthropic's Dario Amodei and OpenAI's Sam Altman both recently talked publicly about pacing the frontier — which we covered in the AI labs' slowdown talk and OpenAI's $1.5 trillion paradox. A few weeks later, both shipped new models on the same Tuesday.

In fairness to both, cheaper and faster isn't necessarily a step up the frontier — Opus 5.5 is positioned as matching an existing tier at lower cost, not exceeding it. But the release pace hasn't visibly slowed, and the competition is now as much about price as capability.