xAI Grok 4.6 Model Card - Factuality (Hallucination): leaderboard
xAI's internal factuality benchmark measures how often a model asserts unsupported or fabricated claims in a single response to an information-seeking query, as flagged by a separate grader against retrieved or known ground truth. xAI Grok 4.6 model card (revised 2026-08-17), section 6.1 chart; hallucination rate in percent, lower is better.
Metric: Hallucination Rate (%, lower is better). Source: media.x.ai. Saturation forecast: Estimated already saturated. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4.5 (High) | 0.98 | #54 (Grok 4.5) |
| 2 | GPT-5.5 (xHigh) | 1.1 | #33 (GPT-5.5) |
| 3 | Grok 4.6 (High) | 1.7 | #29 (Grok 4.6) |
| 4 | Claude Opus 4.8 (Max) | 3.4 | #39 (Claude Opus 4.8) |
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-factuality-hallucination · How It Works · Data refreshed daily, snapshot 2026-11-05.