Phi-3.5-MoE-instruct: benchmark results
Microsoft's 42B mixture-of-experts Phi, activating 6.6B parameters across 16 experts, with a 128K context under MIT license (August 2024). Provider: Microsoft. Released 2024-08-23. Access: Open.
Aggregate Rating 1525 ± 16, rank #700 of 1656 rated models, from 133 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Stats (Social IQa) | 78 | Score (%) | 100 |
| PIQA | 88.6 | Accuracy (%) | 96.7 |
| Open CoT - LogiQA 2 | 15.14 | CoT Gain (%) | 94.7 |
| AILuminate Safety | 4 | Safety Grade (1-5) | 91.9 |
| GSM8K | 88.7 | Accuracy (%) | 91.9 |
| Open CoT - LogiQA | 8.47 | CoT Gain (%) | 91.6 |
| Enkrypt AI - Safety Risk | 18.44 | Risk Score | 91.5 |
| Open LLM Leaderboard - GPQA | 35.57 | Score | 89.6 |
| Open LLM Leaderboard - BBH | 64.08 | Score | 88.8 |
| Open CoT Leaderboard | 13.08 | Average CoT Gain (%) | 87 |
| Open PL LLM - PSC (generative, 5-shot) | 97.62 | Binary F1 (%) | 86.6 |
| Open LLM Leaderboard - MuSR | 45.65 | Score | 86.4 |
Interactive version: theaggregate.ai/model?slug=phi-3-5-moe-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.