Grok 4.20 0309 (Reasoning): benchmark results
March 9, 2026 Grok 4.20 snapshot evaluated with reasoning enabled. Provider: xAI. Released 2026-03-09. Access: API.
Aggregate Rating 1708 ± 10, rank #332 of 2168 rated models, from 230 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA IFBench | 82.93 | Accuracy (%) | 99.8 |
| UGI Leaderboard | 64.23 | UGI Score | 99.4 |
| Nejumi 4 - HLE (Japanese) - Calibration Error | 39 | Calibration error (%) | 99 |
| Nejumi 4 - BFCL - Relevance Detection | 100 | Accuracy (%) | 98.6 |
| AA-Omniscience Hallucination Rate | 22.4 | Hallucination Rate (%) | 97.4 |
| Tau2-Bench Telecom | 96.49 | Success Rate (%) | 97.2 |
| AA TAU-2 Bench | 96.49 | Accuracy (%) | 97 |
| AA Global-MMLU-Lite - Burmese | 88.83 | Accuracy (%) | 95.5 |
| AIME | 96.46 | Score (%) | 94.7 |
| Vals AI AIME | 96.46 | Accuracy (%) | 94.7 |
| AA Global-MMLU-Lite - Italian | 92.75 | Accuracy (%) | 94.5 |
| Nejumi 4 - ARC-AGI-1 | 98 | Accuracy (%) | 93.4 |
Interactive version: theaggregate.ai/model?slug=grok-4-20-0309-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.