Model comparison
GPT-4o vs Phi-4
Phi-4 is the stronger model overall, scoring 30.9 to 28.5 on the Noometry Index.
Last verified . 36 shared benchmarks.
MicrosoftSummary
- They share 36 benchmarks with published results for both. GPT-4o scores higher in 4 categories and Phi-4 in 5 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where GPT-4o leads 52.8 to 40.7.
- The biggest single-benchmark swing is LiveBench Language: 47.6% for GPT-4o and 25.6% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $2.50 / $10 for GPT-4o.
- Phi-4 has downloadable open weights; the other is API-only.
Side by side
| GPT-4o | Phi-4 | |
|---|---|---|
| Provider | OpenAI | Microsoft |
| Noometry Index | 28.5 | 30.9 |
| Released | 2024-05-13 | 2024-12-11 |
| Weights | Proprietary | Open |
| Context window | 128K | 128K |
| Max output | 16K | 4K |
| Input $ / M tokens | $2.50 | $0.07 |
| Output $ / M tokens | $10 | $0.14 |
| Results tracked | 72 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Phi-4 leads
GPT-4o: 25.1 (#323), Phi-4: 33.8 (#237)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| BigCodeBench Instruct | 51.1% | 45.5% |
| LiveBench Coding | 51.4% | 30.7% |
| LMArena Coding | 1297 | 1231 |
| BigCodeBench Complete | 61.1% | 55.4% |
| SWE-bench Verified | 31% | — |
| SWE-bench Verified (bash only) | 21.6% | — |
| Aider Polyglot | 45.3% | — |
| GSO | 0% | — |
| WeirdML | 25.1% | — |
| CadEval | 26% | — |
| HumanEval+ | 87.2% | — |
| MBPP+ | 72.2% | — |
Agentic & Tool Use Phi-4 leads
GPT-4o: 20.8 (#133), Phi-4: 22.6 (#120)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| BALROG | 32.3% | 11.6% |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| GDPval | 9.9% | — |
| TheAgentCompany | 8.6% | — |
| Cybench | 12.5% | — |
| LMArena Search | 1006 | — |
| METR Time Horizons | 40.8% | — |
Reasoning Phi-4 leads
GPT-4o: 9.4 (#338), Phi-4: 17.0 (#287)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| Chess Puzzles | 13% | 1% |
| LiveBench Reasoning | 55.8% | 47.8% |
| LMArena Hard Prompts | 1281 | 1220 |
| LiveBench Data Analysis | 60.9% | 45.2% |
| Epoch Capabilities Index | 128.97 | 130.43 |
| LiveBench | 55.3% | 41.6% |
| ARC-AGI-2 | 0% | — |
| SimpleBench | 17.8% | — |
| ARC-AGI-1 | 4.5% | — |
| CritPt | 0% | — |
| EnigmaEval | 0.8% | — |
| DTBench | 64.5% | — |
| LMCA | 16.6% | — |
| ForecastBench | 57.7 | — |
Math Phi-4 leads
GPT-4o: 10.2 (#303), Phi-4: 19.6 (#279)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 6.4% | 13.8% |
| LiveBench Math | 49.5% | 42% |
| LMArena Math | 1285 | 1246 |
| MATH Level 5 | 53.3% | 64.9% |
| FrontierMath (Tiers 1-3) | 0.4% | — |
| Omni-MATH | 29.3% | — |
| FrontierMath (Feb 2025 set) | 0.3% | — |
Knowledge Phi-4 leads
GPT-4o: 28.8 (#235), Phi-4: 32.5 (#204)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| GPQA Diamond | 49.2% | 56.1% |
| Confabulations | 15.3% | 29.4% |
| Vectara Hallucination Rate | 9.6% | 3.7% |
| LMArena Expert | 1250 | 1203 |
| MMLU | 88.1% | 84.8% |
| Humanity's Last Exam | 2.7% | — |
| SimpleQA Verified | 26% | — |
| MMLU-Pro | 71.3% | — |
| GPQA (HELM) | 52% | — |
Multimodal Not comparable
GPT-4o: 33.9 (#85), Phi-4: —
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| LMArena Vision | 1137 | — |
| Video-MME | 71.9% | — |
| GeoBench | 71% | — |
| VPCT | 40% | — |
| ScienceQA | 88.5% | — |
Multilingual GPT-4o leads
GPT-4o: 43.4 (#181), Phi-4: 37.4 (#232)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| LMArena Non-English | 1283 | 1197 |
| LMArena Chinese | 1277 | 1211 |
| LMArena French | 1303 | 1223 |
| LMArena German | 1282 | 1222 |
| LMArena Japanese | 1257 | 1158 |
| LMArena Korean | 1234 | 1151 |
| LMArena Russian | 1286 | 1209 |
| LMArena Spanish | 1292 | 1234 |
Instruction Following GPT-4o leads
GPT-4o: 66.8 (#202), Phi-4: 60.6 (#246)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| LiveBench Instruction Following | 68.6% | 58.4% |
| LMArena Instruction Following | 1278 | 1201 |
| IFEval | 81.7% | — |
Long Context GPT-4o leads
GPT-4o: 39.4 (#174), Phi-4: 36.9 (#221)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1289 | 1217 |
| Fiction.LiveBench | 66.7% | — |
Writing & Preference GPT-4o leads
GPT-4o: 52.8 (#160), Phi-4: 40.7 (#239)
| Benchmark | GPT-4o | Phi-4 |
|---|---|---|
| LMArena Text | 1300 | 1217 |
| LMArena Creative Writing | 1291 | 1182 |
| Short-Story Creative Writing | 81.8% | 62.6% |
| LMArena Multi-Turn | 1302 | 1206 |
| LiveBench Language | 47.6% | 25.6% |
| WildBench | 82.8% | — |
Frequently asked questions
Is GPT-4o better than Phi-4?
Phi-4 is the stronger model overall, scoring 30.9 to 28.5 on the Noometry Index.
Which is cheaper, GPT-4o or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; GPT-4o lists at $2.50 and $10.
Is GPT-4o or Phi-4 better for coding?
Phi-4 scores higher on coding benchmarks: 33.8 versus 25.1 in the Noometry coding category.
Which has the bigger context window?
Both accept 128K tokens.
How many benchmarks do GPT-4o and Phi-4 share?
36 benchmarks have published results for both models. GPT-4o has 72 scored results on Noometry and Phi-4 has 37.