Model comparison
GPT-4 vs Phi-2
GPT-4 has enough public results to be ranked (#308); Phi-2 does not yet, so treat this comparison as directional.
Last verified . 7 shared benchmarks.
MicrosoftSummary
- They share 7 benchmarks with published results for both.
- Phi-2 has downloadable open weights; the other is API-only.
Side by side
| GPT-4 | Phi-2 | |
|---|---|---|
| Provider | OpenAI | Microsoft |
| Noometry Index | 28.9 | — |
| Released | 2023-03-14 | 2023-12-12 |
| Weights | Proprietary | Open |
| Context window | 8K | — |
| Max output | 8K | — |
| Input $ / M tokens | $30 | — |
| Output $ / M tokens | $60 | — |
| Results tracked | 38 | 11 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Not comparable
GPT-4: 31.2 (#277), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| HumanEval+ | 79.3% | 45.1% |
| WeirdML | 12.4% | — |
| BigCodeBench Instruct | 46% | — |
| LMArena Coding | 1254 | — |
| BigCodeBench Complete | 57.2% | — |
| MBPP+ | — | 54.2% |
Agentic & Tool Use Not comparable
GPT-4: —, Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| METR Time Horizons | 36.1% | — |
Reasoning Not comparable
GPT-4: 17.3 (#285), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| BIG-Bench Hard | 75.1% | 59.4% |
| Epoch Capabilities Index | 125.88 | 107.62 |
| HellaSwag | 95.3% | 53.6% |
| WinoGrande | 87.5% | 54.7% |
| Chess Puzzles | 4% | — |
| LMArena Hard Prompts | 1241 | — |
| Mystery Game Puzzles | 12% | — |
| DTBench | 62.7% | — |
| LMCA | 17.1% | — |
| Adversarial NLI | — | 42.5% |
| ForecastBench | 57.8 | — |
Math Not comparable
GPT-4: 10.1 (#304), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | — |
| LMArena Math | 1269 | — |
| MATH Level 5 | 23% | — |
| GSM8K | 92% | — |
Knowledge Not comparable
GPT-4: 18.4 (#277), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| MMLU | 86.4% | 58.4% |
| TriviaQA | 84.8% | 45.2% |
| GPQA Diamond | 35.7% | — |
| LMArena Expert | 1211 | — |
| ARC (AI2) Challenge | — | 75.9% |
| OpenBookQA | — | 73.6% |
Multilingual Not comparable
GPT-4: 40.8 (#210), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| LMArena Non-English | 1246 | — |
| LMArena Chinese | 1242 | — |
| LMArena French | 1282 | — |
| LMArena German | 1251 | — |
| LMArena Japanese | 1209 | — |
| LMArena Korean | 1184 | — |
| LMArena Russian | 1251 | — |
| LMArena Spanish | 1261 | — |
Instruction Following Not comparable
GPT-4: 65.5 (#217), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| LMArena Instruction Following | 1241 | — |
Long Context Not comparable
GPT-4: 37.7 (#207), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| LMArena Longer Query | 1244 | — |
Writing & Preference Not comparable
GPT-4: 35.0 (#263), Phi-2: —
| Benchmark | GPT-4 | Phi-2 |
|---|---|---|
| LMArena Text | 1263 | — |
| LMArena Creative Writing | 1244 | — |
| EQ-Bench Creative Writing | 752 | — |
| LMArena Multi-Turn | 1257 | — |
Frequently asked questions
Is GPT-4 better than Phi-2?
GPT-4 has enough public results to be ranked (#308); Phi-2 does not yet, so treat this comparison as directional.
How many benchmarks do GPT-4 and Phi-2 share?
7 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and Phi-2 has 11.