tech
Vibe Check: Opus 4.8—Anthropic Should’ve Rounded Up to 5
Opus 4.8 tops both our Senior Engineer benchmark and our writing tests. It’s the most complete model we’ve tested. We just wish it had an app to match.
TL;DR
- Opus 4.8 is a significant advancement over Opus 4.7, excelling in coding, writing, and knowledge work.
- It outperforms GPT-5.5 on the Senior Engineer benchmark (63/100 vs 62/100).
- The model's performance, especially in writing, is heavily reliant on the selected effort level ('extra-high' is competitive, 'high' is adequate, 'medium' is subpar).
- Opus 4.8 demonstrates strong performance in writing, scoring 79.6 on the writing benchmark at high effort.
- The Claude desktop app is described as a 'mess,' with a disjointed user interface that hinders efficient use of the powerful Opus 4.8 model.
- The model is capable of handling large contexts, such as entire codebases or book-length manuscripts.
- Despite its power, Opus 4.8 is slower than GPT-5.5 and can sometimes be too verbose in comments.
- The 'extra-high' effort level is recommended for Opus 4.8 to achieve its full potential as a competitive senior engineer.