AssertLLM2 - Proof Coverage: leaderboard
Proof coverage (%), the share of design logic that the formal solver actually needs to prove the generated assertions on AssertLLM2 (2026), where LLMs write SystemVerilog assertions from the specifications of 83 real-world hardware designs, checked by formal property verification; mean of three generations per design. Higher is better.
Metric: Proof coverage (%): share of the design logic the formal solver needs to prove the generated assertions, SystemVerilog assertions generated from the structured specification of each of 83 AssertLLM2 designs and checked in Cadence JasperGold, mean of three independent generations per design (the paper's default average setting); higher is better. Source: arxiv.org. Saturation forecast: Around August 2027. 6 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 23.3 | #267 |
| 2 | GPT-5.2 | 22.46 | #190 |
| 3 | Gemini 2.5 Pro | 20.19 | #280 |
| 4 | DeepSeek V3.2 | 16.75 | #425 |
| 5 | Qwen 3 Coder Plus | 12.77 | #729 |
| 6 | Llama 4 Maverick | 9.09 | #1113 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=assertllm2-proof-coverage · How It Works · Data refreshed daily, snapshot 2026-11-22.