Grok 4.3 (Low): benchmark results

Provider: xAI. Released 2026-04-30. Access: API.

Unified ELO 1645 ± 1, rank #261 of 1761 rated models, from 18 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA IFBench80.95Accuracy (%)98.7
CritPt60Accuracy (self-reported)91.9
AA Omniscience13.93Score89.1
AA TAU-2 Bench88.89Accuracy (%)83.3
Artificial Analysis Intelligence Index28.53Intelligence Index78.8
AA GPQA Diamond84.34Accuracy (%)75.9
AA Long Context Reasoning74Accuracy (%)75.6
AA Omniscience - Software Engineering (SWE)39.3Accuracy (%)71
AA Humanity's Last Exam18.35Accuracy (%)69.7
AA Omniscience - Science, Engineering & Mathematics34.9Accuracy (%)69.3
AA Omniscience - Business22.2Accuracy (%)68.9
AA Terminal-Bench Hard26.52Accuracy (%)68.1

Interactive version: theaggregate.ai/model?slug=grok-4-3-low · How It Works · Data refreshed daily, snapshot 2026-09-05.