Model comparison
GPT-4o vs gpt-oss-120b
gpt-oss-120b is the stronger model overall, scoring 36.0 to 28.5 on the Noometry Index.
Last verified . 38 shared benchmarks.
Summary
- They share 38 benchmarks with published results for both. GPT-4o scores higher in 3 categories and gpt-oss-120b in 6 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where gpt-oss-120b leads 50.6 to 10.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 6.4% for GPT-4o and 88.9% for gpt-oss-120b.
- gpt-oss-120b is cheaper at $0.037 / $0.17 per million input/output tokens, against $2.50 / $10 for GPT-4o.
- gpt-oss-120b accepts more context: 131K tokens versus 128K.
- gpt-oss-120b has downloadable open weights; the other is API-only.
Side by side
| GPT-4o | gpt-oss-120b | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 28.5 | 36.0 |
| Released | 2024-05-13 | 2025-08-05 |
| Weights | Proprietary | Open |
| Context window | 128K | 131K |
| Max output | 16K | 41K |
| Input $ / M tokens | $2.50 | $0.037 |
| Output $ / M tokens | $10 | $0.17 |
| Results tracked | 72 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding gpt-oss-120b leads
GPT-4o: 25.1 (#323), gpt-oss-120b: 33.2 (#251)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| SWE-bench Verified (bash only) | 21.6% | 26% |
| Aider Polyglot | 45.3% | 41.8% |
| WeirdML | 25.1% | 48.2% |
| LMArena Coding | 1297 | 1380 |
| SWE-bench Verified | 31% | — |
| SciCode | — | 36% |
| GSO | 0% | — |
| BigCodeBench Instruct | 51.1% | — |
| LiveBench Coding | 51.4% | — |
| BigCodeBench Complete | 61.1% | — |
| CadEval | 26% | — |
| ALE-Bench | — | 575.62 |
| AlgoTune | — | 1.41 |
| HumanEval+ | 87.2% | — |
| MBPP+ | 72.2% | — |
Agentic & Tool Use GPT-4o leads
GPT-4o: 20.8 (#133), gpt-oss-120b: 12.1 (#145)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| METR Time Horizons | 40.8% | 56.6% |
| Terminal-Bench | — | 18.7% |
| APEX-Agents | — | 4.4% |
| GDPval | 9.9% | — |
| TheAgentCompany | 8.6% | — |
| Cybench | 12.5% | — |
| BALROG | 32.3% | — |
| LMArena Search | 1006 | — |
| Vending-Bench 2 | — | -21.53 |
Reasoning gpt-oss-120b leads
GPT-4o: 9.4 (#338), gpt-oss-120b: 19.5 (#236)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| SimpleBench | 17.8% | 22.1% |
| CritPt | 0% | 1.1% |
| Chess Puzzles | 13% | 20% |
| LMArena Hard Prompts | 1281 | 1364 |
| DTBench | 64.5% | 76.3% |
| LMCA | 16.6% | 22.1% |
| Epoch Capabilities Index | 128.97 | 139.94 |
| ARC-AGI-2 | 0% | — |
| Kagi LLM Benchmark | — | 58.6% |
| ARC-AGI-1 | 4.5% | — |
| EnigmaEval | 0.8% | — |
| LiveBench Reasoning | 55.8% | — |
| Mystery Game Puzzles | — | 2% |
| LiveBench Data Analysis | 60.9% | — |
| Surface Evolver Bench | — | 25% |
| ForecastBench | 57.7 | — |
| LiveBench | 55.3% | — |
Math gpt-oss-120b leads
GPT-4o: 10.2 (#303), gpt-oss-120b: 50.6 (#43)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 6.4% | 88.9% |
| Omni-MATH | 29.3% | 68.8% |
| LMArena Math | 1285 | 1388 |
| FrontierMath (Tiers 1-3) | 0.4% | — |
| LiveBench Math | 49.5% | — |
| MATH Level 5 | 53.3% | — |
| FrontierMath (Feb 2025 set) | 0.3% | — |
Knowledge gpt-oss-120b leads
GPT-4o: 28.8 (#235), gpt-oss-120b: 42.3 (#89)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| GPQA Diamond | 49.2% | 75.8% |
| MMLU-Pro | 71.3% | 79.5% |
| Confabulations | 15.3% | 15.7% |
| Vectara Hallucination Rate | 9.6% | 14.2% |
| GPQA (HELM) | 52% | 68.4% |
| LMArena Expert | 1250 | 1356 |
| Humanity's Last Exam | 2.7% | — |
| SimpleQA Verified | 26% | — |
| MMLU | 88.1% | — |
Multimodal Not comparable
GPT-4o: 33.9 (#85), gpt-oss-120b: —
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| LMArena Vision | 1137 | — |
| Video-MME | 71.9% | — |
| GeoBench | 71% | — |
| VPCT | 40% | — |
| ScienceQA | 88.5% | — |
Multilingual gpt-oss-120b leads
GPT-4o: 43.4 (#181), gpt-oss-120b: 48.2 (#142)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| LMArena Non-English | 1283 | 1350 |
| LMArena Chinese | 1277 | 1384 |
| LMArena French | 1303 | 1370 |
| LMArena German | 1282 | 1355 |
| LMArena Japanese | 1257 | 1332 |
| LMArena Korean | 1234 | 1283 |
| LMArena Russian | 1286 | 1343 |
| LMArena Spanish | 1292 | 1388 |
Instruction Following gpt-oss-120b leads
GPT-4o: 66.8 (#202), gpt-oss-120b: 69.5 (#168)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| IFEval | 81.7% | 83.6% |
| LMArena Instruction Following | 1278 | 1318 |
| LiveBench Instruction Following | 68.6% | — |
Long Context GPT-4o leads
GPT-4o: 39.4 (#174), gpt-oss-120b: 31.4 (#273)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| Fiction.LiveBench | 66.7% | 44.4% |
| LMArena Longer Query | 1289 | 1319 |
Writing & Preference GPT-4o leads
GPT-4o: 52.8 (#160), gpt-oss-120b: 46.7 (#213)
| Benchmark | GPT-4o | gpt-oss-120b |
|---|---|---|
| LMArena Text | 1300 | 1366 |
| LMArena Creative Writing | 1291 | 1274 |
| Short-Story Creative Writing | 81.8% | 77.1% |
| WildBench | 82.8% | 84.5% |
| LMArena Multi-Turn | 1302 | 1340 |
| EQ-Bench Creative Writing | — | 961 |
| LiveBench Language | 47.6% | — |
Frequently asked questions
Is GPT-4o better than gpt-oss-120b?
gpt-oss-120b is the stronger model overall, scoring 36.0 to 28.5 on the Noometry Index.
Which is cheaper, GPT-4o or gpt-oss-120b?
gpt-oss-120b is cheaper. It lists at $0.037 per million input tokens and $0.17 per million output tokens; GPT-4o lists at $2.50 and $10.
Is GPT-4o or gpt-oss-120b better for coding?
gpt-oss-120b scores higher on coding benchmarks: 33.2 versus 25.1 in the Noometry coding category.
Which has the bigger context window?
gpt-oss-120b does, with 131K tokens against 128K.
How many benchmarks do GPT-4o and gpt-oss-120b share?
38 benchmarks have published results for both models. GPT-4o has 72 scored results on Noometry and gpt-oss-120b has 48.