xAI Grok 4.6 Model Card - PartBench: leaderboard
PartBench evaluates agents on parametric mechanical part design and CAD-style generation, graded for geometric and manufacturing fidelity. xAI Grok 4.6 model card (revised 2026-08-17), section 4.3 chart; the evaluations were conducted by Normal Factory. Reward in percent, higher is better; Grok 4.6 in Grok Build, peers in their providers' harnesses.
Metric: Reward (%). Source: media.x.ai. Saturation forecast: Around May 2028. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 (Max) | 59.9 | #20 (Claude Fable 5) |
| 2 | GPT-5.6 Sol (xHigh) | 59 | #22 (GPT-5.6 Sol) |
| 3 | Grok 4.6 (xHigh) | 58.5 | #29 (Grok 4.6) |
| 4 | Claude Opus 5 (Max) | 52.4 | #16 (Claude Opus 5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=xai-grok-4-6-model-card-partbench · How It Works · Data refreshed daily, snapshot 2026-11-05.