Anthropic’s Claude Opus 4.5 Sets New Bar for Coding AI

Anthropic on Monday announced the release of Claude Opus 4.5, calling it its newest flagship AI model. The company says the model is designed to handle complex technical work, everyday office tasks, and in-depth research.

In its blog, the company described the model as “intelligent, efficient, and the best model in the world for coding, agents, and computer use.” It also said the upgrade marks a meaningful shift in “everyday tasks like deep research and working with slides and spreadsheets.”

This marks the company’s third major release in two months, following Sonnet 4.5 and Haiku 4.5.

Early testing: ‘It just gets it,’ Anthropic says

Inside Anthropic, internal testers consistently offered positive feedback. The company wrote that Claude Opus 4.5 “just ‘gets it,’” noting how it handles ambiguity, navigates tradeoffs, and diagnoses messy, multi-part bugs with minimal guidance.

Anthropic also put the model through a notoriously tough two-hour engineering exam normally given to prospective hires. According to Anthropic, Opus 4.5 “scored higher than any human candidate ever.” The test focuses purely on technical problem-solving, not soft skills. 

Still, the result underscores the company’s argument that advanced AI is starting to rival or surpass expert engineers on specific tasks.

Across nearly all major coding benchmarks, Opus 4.5 hits state-of-the-art marks. On SWE-Bench Verified, a widely watched dataset, the model became the first to surpass 80%, edging out Google’s Gemini 3 Pro and OpenAI’s GPT-5.1 in some evaluations.

The model also comes with reduced pricing: $5 per million input tokens and $25 per million output tokens, a large drop from previous Opus rates of $15 and $75 per million tokens (input/output).

New tools, new integrations

Anthropic is also pairing Opus 4.5 with a round of product upgrades designed to showcase the model’s enhanced “computer use” capabilities. Claude for Chrome is now rolling out to all Max subscribers, giving the system a wider reach within the browser. Claude for Excel is also expanding, becoming available to Max, Team, and Enterprise customers. 

The company’s coding environment, Claude Code, is introducing a more deliberate “Plan Mode” and will now run directly within Anthropic’s desktop app, enabling users to manage multiple coding sessions simultaneously.

A shot at the top tier

The release positions Anthropic squarely against Google’s Gemini 3 and OpenAI’s GPT-5.1, all of which arrived within a few weeks of one another. Unlike its competitors, Anthropic continues to emphasize enterprise productivity over creative media generation, focusing its strengths on coding, spreadsheets, research, and agentic automation.

Anthropic says Opus 4.5 is now the default model for Pro, Max, and Enterprise customers. With lower prices, expanded integrations, and a strong early technical showing, the company appears intent on cementing its position as the go-to AI for business and engineering work.

Elsewhere, GPT-5.1 Codex Max is positioned as a high-end coding model, offering faster iteration loops and tighter integration with the broader ChatGPT workspace.