Key Info
An OpenCode maintainer stated that a recent benchmark correction was not caused by OpenCode, which he says works correctly out of the box, but by problems in the evaluation harness itself.
Highlights
- The maintainer says the fix was applied on the benchmark/harness side, not in OpenCode.
- He claims OpenCode works correctly without special adjustments.
- He adds he would not be surprised if scores for other agents also need correction.
- No specific numbers or revised leaderboard values were provided in the post.