HWE-Bench (Claude Code): leaderboard
HWE-bench (PKU) hardware-engineering split run with the Claude Code agent: 417 real RTL bug-repair cases from OpenTitan, Ibex, CVA6, Caliptra, XiangShan and Rocket Chip, verified by simulation; percent of cases resolved.
Metric: Resolved (%). Source: pku-liang.github.io. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 (Max) | 74.6 | #51 (Claude Opus 4.7) |
| 2 | Claude Opus 4.6 (High) | 68.1 | #67 (Claude Opus 4.6) |
| 3 | Claude Sonnet 4.6 (High) | 66.7 | #94 (Claude Sonnet 4.6) |
| 4 | DeepSeek V4 Pro (Max) | 60.4 | #112 (DeepSeek V4 Pro) |
| 5 | DeepSeek V4 Flash (Max) | 58.3 | #155 (DeepSeek V4 Flash) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=hwe-bench-claude-code · How It Works · Data refreshed daily, snapshot 2026-11-05.