Grok 4 Fast (Non-reasoning): benchmark results
Grok 4 Fast evaluated with reasoning disabled. Provider: xAI. Released 2025-09-19. Access: API.
Aggregate Rating 1529 ± 10, rank #966 of 2168 rated models, from 187 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Phare - Bias Resistance | 80.26 | Score (%) | 98.5 |
| ForesightSafety Bench (Environmental AI Safety, Jailbreak) - Recycling and Circularity Misconduct | 0 | Attack success rate (%) in the recycling and circularity mis | 90.5 |
| PRISM-VLM - Sycophancy Resistance | 76.4 | Per-axis mean score (%; share of turn-0-correct answers held | 90.2 |
| ShatterMed-QA (Chinese, Easy) | 97.82 | Accuracy (%) on the 2,616 Chinese Easy multiple-choice clini | 82.5 |
| Horangi 4 - Korean Hate Speech | 66 | Accuracy (%) | 82.2 |
| FinMTM - Financial Agent | 39.7 | FinMTM financial agent score (0-100) over 1,000 tool-use tas | 81 |
| UGI Leaderboard | 45.76 | UGI Score | 80.9 |
| UGI - Writing | 43.04 | Writing Score | 80.8 |
| ForesightSafety Bench (Embodied AI Safety) - Living-Being Physical Contact Safety | 52.92 | Violation rate (%) over the living-being physical contact sa | 80 |
| ShatterMed-QA (English, Easy) | 96.96 | Accuracy (%) on the 5,923 English Easy multiple-choice clini | 80 |
| ForesightSafety Bench (Environmental AI Safety, Jailbreak) - Environmental Decision Risks | 2.5 | Attack success rate (%) in the environmental decision risks | 78.6 |
| UGI - Natural Intelligence | 33.1 | NatInt Score | 77.7 |
Interactive version: theaggregate.ai/model?slug=grok-4-fast-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.