Benchmarks / Agentic
CADBench
Uses Seldon's first-party mean verifier score. Score: mean verifier score. Collected by Jake Boggs' benchmark dashboard.
CADBench
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Fable 5 | max | 57.9% |
| 2 | Opus 5 | max | 57.0% |
| 3 | Gemini 3.8 Flash | max | 54.7% |
| 4 | GPT-5.6 Sol | max | 51.2% |
| 5 | Gemini 3.7 Flash | max | 50.1% |
| 6 | GPT-5.6 Terra | max | 25.8% |
| 7 | Muse Spark 1.3 | max | 20.7% |
| 8 | Kimi K3 | max | 17.2% |
| 9 | Qwen 3.8 Max | max | 17.0% |
| 10 | Grok 4.6 | max | 16.7% |
| 11 | Muse Spark 1.2 | max | 13.4% |
| 12 | Inkling | max | 6.8% |