PostEDA-Bench (ReAct) - PPA-Mono: leaderboard
Single-objective PPA set of PostEDA-Bench (Liu et al., University of Minnesota, 2026): a ReAct agent drives an OpenROAD design to a performance, power or area target by retuning flow parameters, or for some performance tasks by restructuring the RTL, while a testbench check guards functional equivalence. Success rate in percent over 35 tasks, each the share of 5 temperature-0 runs that meet the target within 16 agent actions; higher is better.
Metric: Success rate (%; 35 single-objective PPA tasks, mean of 5 temperature-0 runs per task, ReAct agent). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 31B (IT) (Thinking) | 64.56 | #343 (Gemma 4 31B (IT)) |
| 2 | Gemini 3 Flash (Preview) | 48.56 | #141 |
| 3 | GPT-5 Mini | 42.28 | #349 |
| 4 | GPT-5 | 39.42 | #163 |
| 5 | Qwen 3.5 122B A10B | 39.42 | #329 |
| 6 | Qwen 3.5 27B | 30.85 | #358 |
| 7 | Qwen 3.5 9B | 25.71 | #803 |
| 8 | DeepSeek V3.2 | 1.14 | #403 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=posteda-bench-react-ppa-mono · How It Works · Data refreshed daily, snapshot 2026-11-12.