Grok 3 Mini: benchmark results
xAI's lightweight Grok 3 reasoning model that thinks before responding, aimed at cost-efficient math, logic and coding (February 2025). Provider: xAI. Released 2025-02-17. Access: API.
Aggregate Rating 1664 ± 8, rank #281 of 1656 rated models, from 513 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMTU - Table Join | 77.14 | Accuracy (%) | 100 |
| EuroEval Norwegian Common Sense Reasoning | 90.07 | Common Sense Reasoning Average Score (%) | 98.6 |
| AA MATH-500 | 99.2 | Accuracy (%) | 97.5 |
| SycoEval-EM - Acquiescence | 0 | Acquiescence Rate (%) | 97.2 |
| EuroEval Danish Knowledge - Danish Citizen Tests | 98.18 | MCC (x100) | 96.7 |
| Nejumi 4 - jaster (0-shot) - JSICK | 85 | Exact match (%) | 96.2 |
| MMTU - Column Relationship | 67.88 | Accuracy (%) | 96 |
| ProLLM - OpenBook Q&A | 86.7 | Score (%) | 95.7 |
| EuroEval Faroese NLU - ScaLA FO | 34.07 | Matthews correlation coefficient (%) | 95.6 |
| EuroEval Icelandic Common Sense Reasoning | 76.18 | Common Sense Reasoning Average Score (%) | 95.4 |
| EuroEval Italian Knowledge | 83.74 | Knowledge Average Score (%) | 95.4 |
| ProLLM - Function Calling | 92.3 | Score (%) | 95.4 |
Interactive version: theaggregate.ai/model?slug=grok-3-mini · How It Works · Data refreshed daily, snapshot 2026-10-20.