HWE-Bench v1.1 (Codex CLI): leaderboard

HWE-Bench v1.1 (November 2026) tests hardware bug-repair agents on 417 simulation-verified RTL tasks from six projects. Updates affect 208 tasks (188 test scripts and 50 statements, with overlap). The evaluator-container limit is normally 1,800 seconds, with seven 4,500-second task exceptions and five RocketChip test scripts limited to 3,600 seconds; original v1.0 results remain separate. This board uses Codex CLI and retains each model’s native reasoning effort. Resolved tasks as a percentage of 417; higher is better.

Metric: Resolved (%). Source: pku-liang.github.io. Saturation forecast: Estimated already saturated. 3 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-6 AstraMax97.1#6
2GPT-6.1 SolMax97.1#10
3GPT-6 SolMax93.5#26

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=hwe-bench-v1-1-codex-cli · How It Works · Data refreshed daily, snapshot 2026-11-16.