Opus 5's reported win is less about another leaderboard slot and more about the test shape
Humanlayer ran Opus 5, Opus 4.8, and Sonnet 5 through 17 coding checkpoints with requirements revealed one at a time, closer to real agent work than a fully specified prompt.
Anthropic launched a lot of new things and a lot of people still haven't noticed.
The Claude ecosystem continues to evolve rapidly, bringing features that change the way we use Artificial Intelligence in our everyday lives. Among the main releases: 🧠 Record Skill: transform
Spidey runner game based on the Spiderverse movies!
Stack: usecrayon, Threejs and Opus 5
You have reached the end of the archive
All of Claude Opus 5