Opus 4.6 is now available in Kiro

Starting today, you can use Claude Opus 4.6, Anthropic’s newest SoTA model, in the Kiro IDE and CLI. Opus 4.6 is Anthropic’s most powerful model yet and also the world’s best model for coding.

Comparison table of scores for Opus 4.6 over Opus 4.5, Sonnet 4.5, Gemini 3 Pro, and GPT-5.2 (all models). Agentic terminal coding: 65.4%; Agentic coding 80.8%; Agentic computer use 72.7%; Agentic tool use retail 91.9% telecom 99.3%; Scaled tool use 59.5%; Agentic search 84.0%; Multidisciplinary reasoning 40.0% without tools 53.1% with tools; Agentic inancial analysis 60.7%; Office tasks 1606; Novel problem-solving 68.6%; Graduate-level reasoning 91.3%; Visual reasoning 73.9% without tools 77.3% with tools; Multilingual Q&A 91.1%

Opus 4.6 maintains everything you love about 4.5, while expanding its coding capabilities to become the best model for production code and sophisticated agents. It excels on large-scale codebases and long-horizon projects, helping senior engineers complete multi-day projects in hours by delegating complex tasks to Opus 4.6 with less oversight.

We see Opus 4.6 as the strongest model for Kiro’s spec-driven development workflows. With Opus 4.6 under the hood, Kiro is able to create detailed but precise specs on top of large existing projects, making updates with surgical precision with minimal user input.

Opus 4.6 is available with experimental support to all Kiro Pro, Pro+, and Power customers, logging in with Google, GitHub, AWS BuilderID and AWS IAM Identity Center. Opus 4.6 is available in the AWS US-East-1 (Northern Virginia) region, and features a 2.2x credit multiplier, the same as Claude Opus 4.5.

Download Kiro or restart the app or CLI to use the new model.

Refactoring made right: how program analysis makes AI agents safe and reliable