xAI Grok 4.6 Model Card - InferenceEval: leaderboard
xAI's internal InferenceEval: agents land production changes to SpaceXAI's inference stack from a frozen pre-change image, scored by hidden GPU unit tests, an integration probe that zeroes failures, and a contract check. xAI Grok 4.6 model card (revised 2026-08-17), section 5.2 chart; accuracy in percent, higher is better; Grok runs in the Grok Build harness.
Metric: Accuracy (%). Source: media.x.ai. Saturation forecast: Around 2029. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4.6 (High) | 46.9 | #29 (Grok 4.6) |
| 2 | GPT-5.6 Sol (Max) | 44.1 | #22 (GPT-5.6 Sol) |
| 3 | Claude Opus 5 (Max) | 43 | #16 (Claude Opus 5) |
| 4 | Grok 4.5 (High) | 41.3 | #54 (Grok 4.5) |
| 5 | Claude Opus 4.8 (Max) | 40.7 | #39 (Claude Opus 4.8) |
| 6 | Kimi K3 (Max) | 39.8 | #35 (Kimi K3) |
| 7 | GPT-5.5 (xHigh) | 38.9 | #33 (GPT-5.5) |
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-inferenceeval · How It Works · Data refreshed daily, snapshot 2026-11-05.