Anthropic releases Claude Opus 4.1

Background

Claude Opus 4.1 was an upgrade to Claude Opus 4, the model it replaced in Anthropic's lineup.^[1] Anthropic's Claude models reach ordinary users through the Claude apps and reach developers through an API (application programming interface), a standard way for one program to call another.

What happened

On 5 August 2025 Anthropic released Claude Opus 4.1, an upgrade to Claude Opus 4 for agentic tasks, real-world coding and reasoning. Agentic tasks are jobs in which the model carries out a sequence of actions, such as running tools or editing files, instead of answering a single question.^[1]

The model scored 74.5% on SWE-bench Verified, a coding benchmark.^[1] Its pricing was unchanged from Opus 4.^[1]

Key details

The model was available to paid Claude users, in Claude Code, on the API under the identifier claude-opus-4-1-20250805, on Amazon Bedrock and on Google Cloud Vertex AI.^[1]

Customers reported gains over Opus 4. GitHub, Rakuten Group and Windsurf reported improvements, especially in multi-file refactoring, which means changing code that spans several files without changing its behaviour, and in precise debugging.^[1]

Impact

Anthropic said that substantially larger model improvements would follow in the coming weeks.^[1] For readers, the launch is a clear example of a version upgrade: a higher benchmark score and an unchanged price, announced alongside a signal that bigger changes were on the way.^[1]

A benchmark is a fixed set of tasks, and its score is the share of those tasks a model completes correctly. A single score shows how a model did on one test and does not show how it will behave in everyday work, so readers should compare scores with other measures. When a new version keeps the price of the old one, the comparison is simpler: the same tasks and the same per-token prices apply to both versions.^[1]