Claude Opus 4.7 launches with improved agentic coding benchmarks
Anthropic released Claude Opus 4.7, which scores 64.3% on an agentic coding benchmark — between Opus 4.6's 53.4% and Mythos preview's 77.8% — and brings improved instruction following, multimodal support, and memory.
Score breakdown
Developers using agentic coding tools like Cursor or Claude Code should evaluate Opus 4.7 as a potential upgrade, given its measurable benchmark gains over Opus 4.6 and its reduced need for careful prompt engineering.
- 01Claude Opus 4.7 is Anthropic's newly released model, highlighted for coding tasks.
- 02On an agentic coding benchmark, Opus 4.7 scores 64.3%, up from Opus 4.6's 53.4%.
- 03Mythos preview scored 77.8% on the same benchmark, placing Opus 4.7 between the two.
Anthropic released Claude Opus 4.7, and in a video by Matt Wolfe, the model is positioned as the new top-tier option for coding tasks. The most notable improvement is in agentic coding: Opus 4.6 scored 53.4% on the relevant benchmark, while Mythos preview reached 77.8%, and Opus 4.7 lands at 64.3% — a meaningful jump over its predecessor, though short of the Mythos preview ceiling.
Beyond raw benchmark performance, Opus 4.7 introduces improved instruction following, which Wolfe describes as reducing the need for careful prompt engineering that older Claude models required.
Beyond raw benchmark performance, Opus 4.7 introduces improved instruction following, which Wolfe describes as reducing the need for careful prompt engineering that older Claude models required. The release also brings enhanced multimodal support — better understanding of images — and improvements to memory. Wolfe states he will use Opus 4.7 going forward when coding with tools such as Cursor or Claude Code, citing the benchmark results as evidence it is currently the best available model for coding.
Key facts
- 01Claude Opus 4.7 is Anthropic's newly released model, highlighted for coding tasks.
- 02On an agentic coding benchmark, Opus 4.7 scores 64.3%, up from Opus 4.6's 53.4%.
- 03Mythos preview scored 77.8% on the same benchmark, placing Opus 4.7 between the two.
- 04Opus 4.7 features improved instruction following, reducing the need for precise prompt engineering.
- 05The model includes improved multimodal support for better image understanding.
- 06Memory handling has also been improved in this release.
- 07Matt Wolfe says he will use Opus 4.7 for coding via tools like Cursor or Claude Code.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content.