AutoEval-ECov (Agentic) - Average Coverage: leaderboard
Metric: Average coverage (%) of the model's SystemVerilog testbenches over the 156 problems of AutoEval-ECov (VerilogEval turned into testbench coverage tasks following CorrectBench, golden RTL and specification given), a failed simulation counting zero; agentic setting: up to three refinement rounds with simulator and coverage feedback; Pass@1 estimated from five samples; models in instruct mode with their recommended sampling; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 Coder 30B A3B Instruct | 94 | #399 |
| 2 | Qwen 3 235B A22B 2507 Instruct | 92.6 | #291 |
| 3 | Qwen 3 30B A3B 2507 Instruct | 91.6 | #464 |
| 4 | Qwen 2.5 Coder 32B Instruct | 88 | #516 |
| 5 | Qwen 2.5 72B Instruct | 87.6 | #436 |
| 6 | Qwen 2.5 Coder 7B Instruct | 77.5 | #927 |
| 7 | Qwen 3 4B 2507 Instruct | 62.8 | #745 |
| 8 | Llama 4 Maverick | 48.7 | #451 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=autoeval-ecov-agentic-average-coverage · How It Works · Data refreshed daily, snapshot 2026-10-11.