Anthropic launched its newest model, Claude Opus 4.5, putting the company back atop the benchmark rankings for AI software coding.
Opus 4.5 scores over 80% on the widely used SWE-bench, which tests models for software engineering skill. Google’s impressive Gemini 3 Pro, launched last week, briefly held the top score with 76.2%.
Anthropic’s Claude product lead Scott White tells Fast Company that the model has also scored higher than any human on the engineering take-home assignment the company gives to engineering job candidates.
Of course, Opus 4.5 does a lot more than coding. Anthropic says it is also the “best model in the world” for powering AI agents and for operating a computer, and that it’s meaningfully better than other models at tasks like deep research and working with slides and spreadsheets.
Opus 4.5 also notched state-of-the-art (best) scores in several other key benchmarks, including “agentic coding SWE-bench verified,” “agentic tool use T-2 bench,” and “novel problem-solving ARC-AGI-2.”
A major challenge with applying AI in real-world work settings is the model’s ability to deal with complexity and ambiguity. White says Anthropic customers feel that Opus 4.5 is better than earlier models at dealing with uncertainty and handling trade-offs without a lot of hand-holding from human workers.
Enterprise customers are increasingly using Anthropic models for office task automation, financial modeling, and document creation, White says. Fundamental Research Labs reported 20% accuracy improvements and 15% efficiency gains on Excel automation tasks using the new model, he adds.
Anthropic has been on a sprint for the past couple of months, releasing Claude Sonnet 4.5 and Haiku 4.5, as well as new products such as Claude Skills, Claude Code, and industry-specific versions for financial services and life sciences.
Opus 4.5 will become the new default model for subscribers of higher-end plans, and will be available as a drop-down menu option for Pro, Standard, Team, and Enterprise users. It’s also available to developer customers via the company’s application programming interface (API), as well as via the Amazon Bedrock, Google Vertex, and Microsoft Azure clouds.
Anthropic says it’s also extending access to a beta version of the Claude plug-in for Chrome, which has been in limited preview, to all Mac users. The company is also making Claude for Excel available to Mac Team and Enterprise users in beta, expanding beyond its previous invite-only research preview.