Model comparison
Phi-2 vs Phi 3 Mini 4k Instruct
Phi 3 Mini 4k Instruct has enough public results to be ranked (#328); Phi-2 does not yet, so treat this comparison as directional.
Last verified . 10 shared benchmarks.
MicrosoftSummary
- They share 10 benchmarks with published results for both.
Side by side
| Phi-2 | Phi 3 Mini 4k Instruct | |
|---|---|---|
| Provider | Microsoft | Microsoft |
| Noometry Index | — | 27.9 |
| Released | 2023-12-12 | 2024-04-23 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 11 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 26.6 (#323)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| HumanEval+ | 45.1% | 59.1% |
| MBPP+ | 54.2% | 54.2% |
| LiveBench Coding | — | 15.5% |
| LMArena Coding | — | 1093 |
Reasoning Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 14.1 (#328)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| Adversarial NLI | 42.5% | 52.8% |
| BIG-Bench Hard | 59.4% | 71.7% |
| HellaSwag | 53.6% | 76.7% |
| WinoGrande | 54.7% | 70.8% |
| Chess Puzzles | — | 0% |
| LiveBench Reasoning | — | 26.8% |
| LMArena Hard Prompts | — | 1072 |
| LiveBench Data Analysis | — | 34.7% |
| Epoch Capabilities Index | 107.94 | — |
| LiveBench | — | 22.4% |
Math Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 26.6 (#257)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| LiveBench Math | — | 15.7% |
| LMArena Math | — | 1111 |
Knowledge Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 28.5 (#246)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| ARC (AI2) Challenge | 75.9% | 84.9% |
| MMLU | 58.4% | 68.8% |
| OpenBookQA | 73.6% | 88% |
| TriviaQA | 45.2% | 64% |
| LMArena Expert | — | 1045 |
Multilingual Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 26.3 (#280)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Non-English | — | 1021 |
| LMArena Chinese | — | 1021 |
| LMArena French | — | 1076 |
| LMArena German | — | 1044 |
| LMArena Japanese | — | 935 |
| LMArena Korean | — | 905 |
| LMArena Russian | — | 1022 |
| LMArena Spanish | — | 1085 |
Instruction Following Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 47.7 (#303)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| LiveBench Instruction Following | — | 39.1% |
| LMArena Instruction Following | — | 1053 |
Long Context Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 31.7 (#276)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Longer Query | — | 1044 |
Writing & Preference Not comparable
Phi-2: —, Phi 3 Mini 4k Instruct: 27.6 (#300)
| Benchmark | Phi-2 | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Text | — | 1073 |
| LMArena Creative Writing | — | 1037 |
| LMArena Multi-Turn | — | 1018 |
| LiveBench Language | — | 9.2% |
Frequently asked questions
Is Phi-2 better than Phi 3 Mini 4k Instruct?
Phi 3 Mini 4k Instruct has enough public results to be ranked (#328); Phi-2 does not yet, so treat this comparison as directional.
How many benchmarks do Phi-2 and Phi 3 Mini 4k Instruct share?
10 benchmarks have published results for both models. Phi-2 has 11 scored results on Noometry and Phi 3 Mini 4k Instruct has 35.