Phi-3 Mini Instruct 3.8B: benchmark results

Provider: Microsoft. Released 2024-04-23. Access: Open.

Unified ELO 1372 ± 1, rank #1614 of 1919 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA Humanity's Last Exam4.96Accuracy (%)29.5
Artificial Analysis Intelligence Index5.82Intelligence Index13.5
AA GPQA Diamond31.92Accuracy (%)7.8
AA IFBench23.88Accuracy (%)6.2
AA Terminal-Bench Hard0Accuracy (%)5.5
BenchmarkList ECI68.7Capability Index (ECI)4.3
AA TAU-2 Bench0Accuracy (%)2.5

Interactive version: theaggregate.ai/model?slug=phi-3-mini-instruct-3-8b · How It Works · Data refreshed daily, snapshot 2026-09-08.