Grok 4: benchmark results
xAI's flagship Grok 4 reasoning model with native tool use. Provider: xAI. Released 2025-07-09. Access: API.
Aggregate Rating 1707 ± 4, rank #181 of 1656 rated models, from 1675 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ACE-Bench (Azure SDK, MCP Documentation) | 68.7 | Strict pass rate (%) on ACE-Bench (353 Azure SDK coding task | 100 |
| AIME 2024 | 94 | Score (percentage) | 100 |
| AJ-Bench (LLM-as-a-Judge) - FileSystem | 75.7 | F1 (%) on FileSystem data-system tasks (MCPMark; 24 tasks, 1 | 100 |
| AgenticRAGTracer (Comparison) - 2-Hop | 80.7 | Exact match (%) on AgenticRAGTracer's 2-hop parallel compari | 100 |
| Balrog | 43.6 | Score (imported) | 100 |
| CommunityBench - Preference Distribution Kendall Tau | 0.17 | Kendall's Tau (-1 to 1) | 100 |
| Condesion-Bench (April-September 2025 Subset) | 89.68 | Decision Satisfaction Rate (%, from a 0-1 share times 100) o | 100 |
| Condesion-Bench (April-September 2025 Subset) - Condition Satisfaction Rate | 96.56 | Condition Satisfaction Rate (%, from a 0-1 share times 100) | 100 |
| DV-World - DV-Evol Vega-Lite | 50.68 | Overall score (%) on the Vega-Lite DV-Evol tasks (adapt a re | 100 |
| DV-World - DV-Inter | 40.43 | DV-Inter score (%): per-task MLLM rubric score (interaction, | 100 |
| DV-World - DV-Inter Interaction Success | 79.57 | Interaction success rate (%): 0.5 + 0.5 x (resolved turns mi | 100 |
| FLY-EVAL++ - Single-Step Prediction - Field Validity | 68 | S1 D2 field score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=grok-4 · How It Works · Data refreshed daily, snapshot 2026-10-20.