Rule2DRC
Models write KLayout scripts from chip design rules. The evaluator checks whether each script flags the correct violations on every test layout.
The authors rank 12 models on 1,000 public tasks and 100 private tasks. They report scores with and without the KLayout documentation.
- Made by
- Seoul National University and collaborators
- Released
- May 2026
- Kind
- Electronics
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- Public, plus a private test set
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- KLayout
- Refresh
- One fixed set
- Run by
- A host
- In views
- Agents
- Strongest model tested
- Kimi K3 #25 on the capability index
- Last model added
- July 2026
- Scores refreshed
- November 20th, 2026
Kimi K3 leads with 67.9% among models released by July 2026
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
The newest model with a score here came out in July 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 Kimi K3 Docs in context, thinking 67.9%
- 2 GLM-5.2 Docs in context, thinking 56.8%
- 3
MiniMax-M3 Docs in context, thinking 40.6%
- 4
DeepSeek-V4-Pro Docs in context, thinking 39.8%
- 5
Tencent Hy3 Docs in context 32.6%
- 6 gpt-oss-120b Docs in context, reasoning effort medium 32.5%
- 7
Qwen3-235B-A22B-Thinking Jul 2025, Docs in context, thinking 30.2%
- 8 MiMo-V2.5 Docs in context, thinking 26.6%
- 9
Qwen3-235B-A22B-Instruct Jul 2025, Docs in context, instruct (non-reasoning) 18.6%
- 10
Qwen3-30B-A3B-Thinking Jul 2025, Docs in context, thinking, 9.8K output cap 17.9%
- 11 gpt-oss-20b Docs in context, reasoning effort medium 16.9%
- 12
Qwen3-30B-A3B-Instruct Jul 2025, Docs in context, instruct (non-reasoning) 14.1%
Source: Official Rule2DRC leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Kimi K3 | 67.9% | Docs in context, thinking | Jul 2026 | Official leaderboard |
| 2 | GLM-5.2 | 56.8% | Docs in context, thinking | Jun 2026 | Official leaderboard |
| 3 | | 40.6% | Docs in context, thinking | Jun 2026 | Official leaderboard |
| 4 | | 39.8% | Docs in context, thinking | Apr 2026 | Official leaderboard |
| 5 | | 32.6% | Docs in context | Jul 2026 | Official leaderboard |
| 6 | gpt-oss-120b | 32.5% | Docs in context, reasoning effort medium | Aug 2025 | Official leaderboard |
| 7 | | 30.2% | Docs in context, thinking | Jul 2025 | Official leaderboard |
| 8 | MiMo-V2.5 | 26.6% | Docs in context, thinking | Apr 2026 | Official leaderboard |
| 9 | | 18.6% | Docs in context, instruct (non-reasoning) | Jul 2025 | Official leaderboard |
| 10 | | 17.9% | Docs in context, thinking, 9.8K output cap | Jul 2025 | Official leaderboard |
| 11 | gpt-oss-20b | 16.9% | Docs in context, reasoning effort medium | Aug 2025 | Official leaderboard |
| 12 | | 14.1% | Docs in context, instruct (non-reasoning) | Jul 2025 | Official leaderboard |
Rankings by part
We read 3 parts from the benchmark’s own leaderboard: Public error, Private success and Private error. Each ranks models on one part of the test. Kimi K3 leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.
Public error
12 models we track have a score here.
- 1 Kimi K3 1.30%
- 2 GLM-5.2 6.00%
- 3
DeepSeek-V4-Pro ↑1 10.5%
- 4
MiniMax-M3 ↓1 15.5%
- 5
Tencent Hy3 30.3%
- 6
Qwen3-235B-A22B-Thinking Jul 2025 ↑1 36.3%
- 7 gpt-oss-120b ↓1 48.5%
- 8 MiMo-V2.5 57.8%
- 9
Qwen3-30B-A3B-Instruct Jul 2025 ↑3 61.9%
- 10
Qwen3-235B-A22B-Instruct Jul 2025 ↓1 63.4%
Private success
12 models we track have a score here.
- 1 Kimi K3 49%
- 2 GLM-5.2 37%
- 3
MiniMax-M3 26%
- 4
Qwen3-235B-A22B-Thinking Jul 2025 ↑3 22%
- 5 gpt-oss-120b ↑1 21%
- 6
DeepSeek-V4-Pro ↓2 19%
- 7
Tencent Hy3 ↓2 16%
- 8
Qwen3-30B-A3B-Thinking Jul 2025 ↑2 15%
- 9 MiMo-V2.5 ↓1 14%
- 10 gpt-oss-20b ↑1 12%
Private error
12 models we track have a score here.
- 1 Kimi K3 2.00%
- 2 GLM-5.2 4.00%
- 3
DeepSeek-V4-Pro ↑1 7.00%
- 4
MiniMax-M3 ↓1 11%
- 5
Qwen3-235B-A22B-Thinking Jul 2025 ↑2 18%
- 6
Tencent Hy3 ↓1 29%
- 7 gpt-oss-120b ↓1 50%
- 8 MiMo-V2.5 52%
- 9
Qwen3-235B-A22B-Instruct Jul 2025 53%
- 10
Qwen3-30B-A3B-Thinking Jul 2025 54%
More effort adds a median 23 points
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official Rule2DRC leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Kimi K3 | Docs in context, thinking | 67.9% | |||
| GLM-5.2 | Docs in context, thinking | 56.8% | |||
| Kimi K3 | No context, thinking | 45.3% | |||
| Docs in context, thinking | 40.6% | ||||
| Docs in context, thinking | 39.8% | ||||
| GLM-5.2 | No context, thinking | 38.1% | |||
| Docs in context | 32.6% | ||||
| gpt-oss-120b | Docs in context, reasoning effort medium | 32.5% | |||
| Docs in context, thinking | 30.2% | ||||
| MiMo-V2.5 | Docs in context, thinking | 26.6% | |||
| No context, thinking | 24.3% | ||||
| No context, thinking | 18.9% | ||||
| Docs in context, instruct (non-reasoning) | 18.6% | ||||
| Docs in context, thinking, 9.8K output cap | 17.9% | ||||
| gpt-oss-20b | Docs in context, reasoning effort medium | 16.9% | |||
| No context | 15.4% | ||||
| No context, thinking | 14.7% | ||||
| Docs in context, instruct (non-reasoning) | 14.1% | ||||
| gpt-oss-120b | No context, reasoning effort medium, OpenRouter | 9.50% | |||
| No context, instruct (non-reasoning) | 9.20% | ||||
| MiMo-V2.5 | No context, thinking | 7.80% | |||
| No context, instruct (non-reasoning), OpenRouter | 5.50% | ||||
| No context, thinking | 5.50% | ||||
| gpt-oss-20b | No context, reasoning effort medium, OpenRouter | 4.60% |
4 models have held number one on the official leaderboard since its release in May 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Rule2DRC leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Rule2DRC leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Engineering and CAD
- EEBench In use 13 electrical-engineering circuit and PCB design tasks graded by deterministic build checks, simulation and BOM constraints. No LLM judge.
Gemini 4 Carbon 83.9%
- 3DCodeBench In use Agentic procedural 3D modeling via code: author engine-ready 3D assets through software APIs. Graded for executability and shape fidelity.
LongCat-2.5 55.4%
- CADGenBench Only in system cards Agents construct CAD models from design prompts. The generation split is graded for executability and geometric correctness. Claude Fable 5 45.1%
- PartBench Only in system cards Parametric mechanical part design and CAD-style generation, graded for geometric and manufacturing fidelity. Grok 4.6 58.5%
- Razavi-Bench Not in model cards Models explain what happens in analog circuits and why. A language model awards up to four points per answer against a reference and a rubric.
Gemini 4 Carbon 98.3%
- CircuitRubric Not in model cards Models write analog circuit netlists from design requests. A program compares transistor connections and relative device sizes with accepted reference circuits. Claude Opus 4.8 96%
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access Public tasks, plus a private test set.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (12 entries).
- Pass Models scored 12 models we track.
- Pass Newest score The newest is from July 2026, for Kimi K3.
- Note Frontier models None of the top 10 has a score. The strongest tested is Kimi K3 (#25).
- Pass Headroom The best is 67.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on jinukkim.me: 24 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.