Model comparison
gpt-oss-120b vs Phi-4
gpt-oss-120b is the stronger model overall, scoring 36.3 to 31.2 on the Noometry Index.
Last verified . 24 shared benchmarks.
MicrosoftSummary
- They share 24 benchmarks with published results for both. gpt-oss-120b scores higher in 6 categories and Phi-4 in 3 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where gpt-oss-120b leads 52.5 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 88.9% for gpt-oss-120b and 13.8% for Phi-4.
- gpt-oss-120b is cheaper at $0.037 / $0.17 per million input/output tokens, against $0.07 / $0.14 for Phi-4.
- gpt-oss-120b accepts more context: 131K tokens versus 128K.
Side by side
| gpt-oss-120b | Phi-4 | |
|---|---|---|
| Provider | OpenAI | Microsoft |
| Noometry Index | 36.3 | 31.2 |
| Released | 2025-08-05 | 2024-12-11 |
| Weights | Open | Open |
| Context window | 131K | 128K |
| Max output | 41K | 4K |
| Input $ / M tokens | $0.037 | $0.07 |
| Output $ / M tokens | $0.17 | $0.14 |
| Results tracked | 48 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
gpt-oss-120b: 33.5 (#256), Phi-4: 34.4 (#239)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| LMArena Coding | 1380 | 1231 |
| SWE-bench Verified (bash only) | 26% | — |
| Aider Polyglot | 41.8% | — |
| SciCode | 36% | — |
| WeirdML | 48.2% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 575.62 | — |
| AlgoTune | 1.41 | — |
Agentic & Tool Use Phi-4 leads
gpt-oss-120b: 12.2 (#153), Phi-4: 22.8 (#128)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| Terminal-Bench | 18.7% | — |
| APEX-Agents | 4.4% | — |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| BALROG | — | 11.6% |
| METR Time Horizons | 56.6% | — |
| Vending-Bench 2 | -21.53 | — |
Reasoning gpt-oss-120b leads
gpt-oss-120b: 20.0 (#245), Phi-4: 17.7 (#291)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| Chess Puzzles | 20% | 1% |
| LMArena Hard Prompts | 1364 | 1220 |
| Epoch Capabilities Index | 139.93 | 130.42 |
| SimpleBench | 22.1% | — |
| Kagi LLM Benchmark | 58.6% | — |
| CritPt | 1.1% | — |
| LiveBench Reasoning | — | 47.8% |
| Mystery Game Puzzles | 2% | — |
| DTBench | 76.3% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 22.1% | — |
| Surface Evolver Bench | 25% | — |
| LiveBench | — | 41.6% |
Math gpt-oss-120b leads
gpt-oss-120b: 52.5 (#50), Phi-4: 20.8 (#285)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9% | 13.8% |
| LMArena Math | 1389 | 1246 |
| Omni-MATH | 68.8% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
Knowledge gpt-oss-120b leads
gpt-oss-120b: 42.4 (#96), Phi-4: 32.6 (#209)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| GPQA Diamond | 75.8% | 56.1% |
| Confabulations | 15.7% | 29.4% |
| Vectara Hallucination Rate | 14.2% | 3.7% |
| LMArena Expert | 1356 | 1203 |
| MMLU-Pro | 79.5% | — |
| GPQA (HELM) | 68.4% | — |
| MMLU | — | 84.8% |
Multilingual gpt-oss-120b leads
gpt-oss-120b: 48.0 (#147), Phi-4: 37.2 (#237)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| LMArena Non-English | 1351 | 1197 |
| LMArena Chinese | 1385 | 1212 |
| LMArena French | 1369 | 1224 |
| LMArena German | 1353 | 1222 |
| LMArena Japanese | 1331 | 1158 |
| LMArena Korean | 1282 | 1151 |
| LMArena Russian | 1343 | 1209 |
| LMArena Spanish | 1389 | 1234 |
Instruction Following gpt-oss-120b leads
gpt-oss-120b: 69.3 (#173), Phi-4: 60.4 (#251)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1318 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
| IFEval | 83.6% | — |
Long Context Phi-4 leads
gpt-oss-120b: 31.4 (#278), Phi-4: 36.9 (#226)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1319 | 1217 |
| Fiction.LiveBench | 44.4% | — |
Writing & Preference gpt-oss-120b leads
gpt-oss-120b: 46.5 (#217), Phi-4: 40.5 (#244)
| Benchmark | gpt-oss-120b | Phi-4 |
|---|---|---|
| LMArena Text | 1365 | 1217 |
| LMArena Creative Writing | 1275 | 1182 |
| Short-Story Creative Writing | 77.1% | 62.6% |
| LMArena Multi-Turn | 1340 | 1206 |
| EQ-Bench Creative Writing | 961 | — |
| WildBench | 84.5% | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is gpt-oss-120b better than Phi-4?
gpt-oss-120b is the stronger model overall, scoring 36.3 to 31.2 on the Noometry Index.
Which is cheaper, gpt-oss-120b or Phi-4?
gpt-oss-120b is cheaper. It lists at $0.037 per million input tokens and $0.17 per million output tokens; Phi-4 lists at $0.07 and $0.14.
Is gpt-oss-120b or Phi-4 better for coding?
They score almost the same on coding (33.5 vs 34.4); test both on your own repository before choosing.
Which has the bigger context window?
gpt-oss-120b does, with 131K tokens against 128K.
How many benchmarks do gpt-oss-120b and Phi-4 share?
24 benchmarks have published results for both models. gpt-oss-120b has 48 scored results on Noometry and Phi-4 has 37.