Benchmark
Parametric CAD Bench v1
Scores are percentages.
Benchmark sourceScores from Harbor Hub Leaderboards.
- Models measured
- 19
- Scores recorded
- 19
- Sources
- 1
- IQ weight
- 1.17
Score distribution
- 55.8%
- Mean
- 51.8%
- Median
- 28.4%
- Std dev
- 11.9%
- 10th percentile
- 88.8%
- 90th percentile
- 5.8%
- Min
This benchmark has a weight of 1.17 for chat models, measured over 19 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5v1· Harbor Hub Leaderboards | 90.6% |
| 2 | Kimi K3v1· Harbor Hub Leaderboards | 90.4% |
| 3 | Grok 4.6v1· Harbor Hub Leaderboards | 88.8% |
| 4 | GPT-5.6 Solv1· Harbor Hub Leaderboards | 86.5% |
| 5 | Grok 4.5v1· Harbor Hub Leaderboards | 83.7% |
| 6 | GPT 5.5v1· Harbor Hub Leaderboards | 83.2% |
| 7 | Gemini 3.1 Pro Previewv1· Harbor Hub Leaderboards | 79% |
| 8 | Claude Fable 5v1· Harbor Hub Leaderboards | 75.8% |
| 9 | Claude Opus 4.7v1· Harbor Hub Leaderboards | 73.4% |
| 10 | Claude Sonnet 4.6v1· Harbor Hub Leaderboards | 51.8% |
| 11 | Kimi K2.7 Codev1· Harbor Hub Leaderboards | 51.5% |
| 12 | Qwen3 Coderv1· Harbor Hub Leaderboards | 43.7% |
| 13 | DeepSeek V3.2v1· Harbor Hub Leaderboards | 37.9% |
| 14 | Devstral 2 123b Instruct 2512v1· Harbor Hub Leaderboards | 37.8% |
| 15 | Qwen3.6 27Bv1· Harbor Hub Leaderboards | 36% |
| 16 | Claude Haiku 4.5v1· Harbor Hub Leaderboards | 35.8% |
| 17 | Codestral 25.01v1· Harbor Hub Leaderboards | 27.7% |
| 18 | Gemini 3.1 Flash Litev1· Harbor Hub Leaderboards | 24.1% |
| 19 | Llama 3.3 70B Instructv1· Harbor Hub Leaderboards | 9.2% |