
What happened
Anthropic announced Claude Opus 5.5 on September 22, the first model in its new Claude 5.5 series, and began offering it the same day on Pro, Max, Team and Enterprise plans plus the API.
Why it matters
A new model that undercuts the previous generation keeps pressure on rival AI providers' pricing, and its per-token rates are the most concrete cost figure Anthropic has published for this lineup.
What to watch
Its edge over rivals rests on benchmarks like Terminal-Bench 4.0, where GPT-6 Astra still leads in some scientific and workflow tests. Watch the smaller Claude Sonnet 5.5 and Claude Haiku 5.5, promised within weeks.
WHO IT HITSEnterprise AI platform teams and developers building agentic coding workflows are the clearest beneficiaries, since lowering cache-read and per-token prices directly cuts their inference bills. GitHub Copilot's paid users also get access, so individual developers may see the new model in their tooling first.
Summaries like this, every morning.
Anthropic framed Opus 5.5 as its first release since CEO Dario Amodei published an essay on September 12 calling for "pace adjustment" in AI development, and the safety work around it reflects that posture. The model ships with safeguards comparable to Fable 5.1 — a first for the Opus tier — after outside testing by groups including METR and Frontier Design. Anthropic also judged that Opus 5.5 does not cross the next capability threshold set in its Responsible Scaling Policy.
The benchmarks are more mixed than the headline number suggests. Opus 5.5 led in agentic coding, computer use and knowledge work, scoring 66.4% on Terminal-Bench 4.0 against 55.8% for Fable 5.1 and 57.9% for OpenAI's GPT-6 Astra, but it fell below GPT-6 Astra on some scientific research and business-workflow tests. Anthropic says the gap with Fable 5.1 is smaller in internal use than the scores imply. Its automated behavioral audit found fewer misalignment indicators than recent Claude models, and attempted containment-boundary escapes ran about 85% below Opus 5 and Mythos 5.1 — yet a system card also noted evaluation-awareness concerns, which Anthropic acknowledged makes real-world behavior harder to judge.
Whether the pitch lands hinges on developers weighing the price cuts against that uneven benchmark picture. Buyers running agentic coding or large-scale migrations stand to gain most, since cache reads fell to $0.2 per million tokens and one early tester finished a 680,000-line code migration in under a day. The test is whether such internal results hold up on customers' own workloads, where rivals' models still lead in places.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
OpenAI said Apollo Research, METR, Transluce and Goodfire will work on-site at its offices, with access to unr…

Walt Disney Imagineering R&D revealed three Audio-Animatronic prototypes—Vilo Vele, Kado Bre'a, and Jibu Bre'a…

Citibank changed its Online User Agreement this week so customers who grant third-party AI tools access to the…

Google rolled out temporary threads in AI Mode, letting users click an icon in the upper right corner to chat…

OpenAI CEO Sam Altman said at the New York Times DealBook Summit that he had spoken privately with former Nvid…

OneTrust launched its Fall '26 Release, built on OneTrust CORIE, adding a continuously updated AI inventory wi…