RTL-BenchLS - Masked-Content Reasoning: leaderboard
Masked-content reasoning on 420 designs: one block or sub-module instantiation of a Verilog design is masked and the model restores it from the surrounding code and a description of the masked logic, from RTL-BenchLS (2026), a benchmark of formally verified Verilog designs; the score is the functional pass rate in percent, where output must be formally equivalent to the golden design (Cadence LEC or SEC), and higher is better.
Metric: Functional pass rate (%) on masked-content reasoning on 420 designs: one always/assign/case block or sub-module instantiation is masked and the model restores it from the surrounding code and its own description of the masked block; functional pass rate (%; output formally equivalent to the golden design, Cadence Conformal LEC for combinational and JasperGold SEC for sequential designs), temperature 0.2; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 28.2 | #1357 |
| 2 | Claude Haiku 4.5 | 27.5 | #568 |
| 3 | Qwen 3.5 397B A17B | 27.5 | #246 |
| 4 | DeepSeek V3.2 | 27.3 | #406 |
| 5 | Claude Sonnet 4.5 | 26.8 | #250 |
| 6 | GPT-4o | 22.4 | #741 |
| 7 | GPT-4o Mini | 21.9 | #1623 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=rtl-benchls-masked-content-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-13.