Phi-3.5 on MMLU-Pro: score, rank and what it means
Phi-3.5 scores 76.1 on MMLU-Pro, placing it #304 of 600 models in the index. The benchmark average is 73.5 and the leader is GPT-5.2 Pro at 91.6.
Score distribution
The chart shows the top 10 models on MMLU-Pro, with Phi-3.5 highlighted. Longer bars mean higher scores.
Top 10 leaderboard
| # | Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|---|
| 1 | GPT-5.2 Pro | OpenAI | 2000000 | $21.32 | 89.6 |
| 2 | GPT-5.1 Max | OpenAI | 2000000 | $20.50 | 87.9 |
| 3 | Claude Opus 4.5 Pro | Anthropic | 200000 | $20.98 | 89.1 |
| 4 | Claude Opus 4.5 Reasoning | Anthropic | 2000000 | $20.91 | 88.2 |
| 5 | Gemini 3 Pro Reasoning | Google DeepMind | 200000 | $19.69 | 87.5 |
| 6 | GPT-5.1 Reasoning | OpenAI | 400000 | $20.27 | 88.3 |
| 7 | GPT-5.2 | OpenAI | 2000000 | $19.71 | 88.0 |
| 8 | Claude Opus 4.5 Max | Anthropic | 1000000 | $20.90 | 87.8 |
| 9 | GPT-5.2 Vision | OpenAI | 1000000 | $19.03 | 87.4 |
| 10 | Gemini 3 Pro Max | Google DeepMind | 2000000 | $20.57 | 87.9 |
What this score means
On MMLU-Pro, Phi-3.5 sits at #304 of 600. The leader, GPT-5.2 Pro, scores 91.6, so the gap to the top is 15.5 points. The median model in this index scores 76.1, which puts Phi-3.5 above the midpoint.
Microsoft models average 68.2 on this benchmark. Phi-3.5 is 7.9 points above that provider line, so it is a stronger result than the average Microsoft model.
Among open-weight models, the average is 72.1. Phi-3.5 is 4.0 points above that group. This is the more useful comparison if you are shortlisting within the same tier rather than across the whole index.
Read a 76.1 as a relative signal, not an absolute guarantee. Benchmark scores depend on prompt format, evaluation harness and contamination controls. Use this page to narrow the shortlist, then test the model on your own data.
Price context
At $2.81 per million input tokens, Phi-3.5 is below the median input price of $4.00 across the 600 models scored on MMLU-Pro. The models immediately around it in the ranking are shown below so you can compare score and price together.
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Claude Opus 4.5 Nano | Anthropic | 32768 | $9.38 | 73.3 |
| Command R+ Base | Cohere | 65536 | $3.24 | 71.6 |
| GLM-4.5 Base | Zhipu AI | 32768 | $3.99 | 74.9 |
| Mixtral Chat | Mistral AI | 65536 | $3.30 | 72.6 |
| Phi-3.5 | Microsoft | 16384 | $2.81 | 70.9 |
| Auto Turbo | OpenRouter | 65536 | $7.96 | 70.9 |
| Firefunction | Fireworks AI | 16384 | $8.16 | 72.3 |
Other benchmarks for Phi-3.5
| Benchmark | Score | Rank | Percentile | Relative |
|---|---|---|---|---|
| GPQA Diamond | 67.0 | 331 | 45th | |
| SWE-bench Verified | 60.9 | 308 | 49th | |
| HumanEval | 76.2 | 327 | 46th | |
| MATH-500 | 75.4 | 316 | 47th | |
| AIME 2025 | 60.6 | 383 | 36th | |
| LiveCodeBench | 59.8 | 371 | 38th | |
| IFEval | 77.1 | 377 | 37th | |
| TruthfulQA | 74.8 | 336 | 44th | |
| HellaSwag | 82.9 | 323 | 46th | |
| MMMU | 68.8 | 328 | 45th | |
| Arena Elo | 1356.7 | 350 | 42th |
FAQ
What score does Phi-3.5 get on MMLU-Pro?
Phi-3.5 scores 76.1 on MMLU-Pro in this index. The benchmark average is 73.5 and the top score is 91.6.
How does Phi-3.5 rank on MMLU-Pro?
It ranks #304 of 600 models, in the 49th percentile. The leader is GPT-5.2 Pro at 91.6.
Is Phi-3.5 good at MMLU-Pro?
Yes, relative to the index: it is above the median and inside the top half of scores. The result should be read alongside its price and context window.
How is the MMLU-Pro score calculated?
MMLU-Pro is a knowledge benchmark. Scores in this index are modelled estimates on a 0–100 scale (or Elo for Arena), derived from capability tier and benchmark difficulty. They are not official leaderboard results.
What does Phi-3.5 cost?
Phi-3.5 costs $2.81 per million input tokens and $14.99 per million output tokens.