Model comparison
GPT-4 vs gpt-oss-120b
gpt-oss-120b is the stronger model overall, scoring 36.3 to 29.1 on the Noometry Index.
Last verified . 27 shared benchmarks.
Summary
- They share 27 benchmarks with published results for both. GPT-4 scores higher in 1 category and gpt-oss-120b in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where gpt-oss-120b leads 52.5 to 10.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.1% for GPT-4 and 88.9% for gpt-oss-120b.
- gpt-oss-120b is cheaper at $0.037 / $0.17 per million input/output tokens, against $30 / $60 for GPT-4.
- gpt-oss-120b accepts more context: 131K tokens versus 8K.
- gpt-oss-120b has downloadable open weights; the other is API-only.
Side by side
| GPT-4 | gpt-oss-120b | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 29.1 | 36.3 |
| Released | 2023-03-14 | 2025-08-05 |
| Weights | Proprietary | Open |
| Context window | 8K | 131K |
| Max output | 8K | 41K |
| Input $ / M tokens | $30 | $0.037 |
| Output $ / M tokens | $60 | $0.17 |
| Results tracked | 38 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding gpt-oss-120b leads
GPT-4: 31.6 (#283), gpt-oss-120b: 33.5 (#256)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| WeirdML | 12.4% | 48.2% |
| LMArena Coding | 1254 | 1380 |
| SWE-bench Verified (bash only) | — | 26% |
| Aider Polyglot | — | 41.8% |
| SciCode | — | 36% |
| BigCodeBench Instruct | 46% | — |
| BigCodeBench Complete | 57.2% | — |
| ALE-Bench | — | 575.62 |
| AlgoTune | — | 1.41 |
| HumanEval+ | 79.3% | — |
Agentic & Tool Use Not comparable
GPT-4: —, gpt-oss-120b: 12.2 (#153)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| METR Time Horizons | 36.1% | 56.6% |
| Terminal-Bench | — | 18.7% |
| APEX-Agents | — | 4.4% |
| Vending-Bench 2 | — | -21.53 |
Reasoning gpt-oss-120b leads
GPT-4: 17.8 (#289), gpt-oss-120b: 20.0 (#245)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| Chess Puzzles | 4% | 20% |
| LMArena Hard Prompts | 1241 | 1364 |
| Mystery Game Puzzles | 12% | 2% |
| DTBench | 62.7% | 76.3% |
| LMCA | 17.1% | 22.1% |
| Epoch Capabilities Index | 125.89 | 139.93 |
| SimpleBench | — | 22.1% |
| Kagi LLM Benchmark | — | 58.6% |
| CritPt | — | 1.1% |
| Surface Evolver Bench | — | 25% |
| BIG-Bench Hard | 75.1% | — |
| ForecastBench | 57.8 | — |
| HellaSwag | 95.3% | — |
| WinoGrande | 87.5% | — |
Math gpt-oss-120b leads
GPT-4: 10.8 (#309), gpt-oss-120b: 52.5 (#50)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 88.9% |
| LMArena Math | 1269 | 1389 |
| Omni-MATH | — | 68.8% |
| MATH Level 5 | 23% | — |
| GSM8K | 92% | — |
Knowledge gpt-oss-120b leads
GPT-4: 18.4 (#282), gpt-oss-120b: 42.4 (#96)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| GPQA Diamond | 35.7% | 75.8% |
| LMArena Expert | 1211 | 1356 |
| MMLU-Pro | — | 79.5% |
| Confabulations | — | 15.7% |
| Vectara Hallucination Rate | — | 14.2% |
| GPQA (HELM) | — | 68.4% |
| MMLU | 86.4% | — |
| TriviaQA | 84.8% | — |
Multilingual gpt-oss-120b leads
GPT-4: 40.6 (#215), gpt-oss-120b: 48.0 (#147)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| LMArena Non-English | 1246 | 1351 |
| LMArena Chinese | 1242 | 1385 |
| LMArena French | 1283 | 1369 |
| LMArena German | 1251 | 1353 |
| LMArena Japanese | 1209 | 1331 |
| LMArena Korean | 1184 | 1282 |
| LMArena Russian | 1251 | 1343 |
| LMArena Spanish | 1261 | 1389 |
Instruction Following gpt-oss-120b leads
GPT-4: 65.3 (#222), gpt-oss-120b: 69.3 (#173)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| LMArena Instruction Following | 1241 | 1318 |
| IFEval | — | 83.6% |
Long Context GPT-4 leads
GPT-4: 37.7 (#212), gpt-oss-120b: 31.4 (#278)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| LMArena Longer Query | 1244 | 1319 |
| Fiction.LiveBench | — | 44.4% |
Writing & Preference gpt-oss-120b leads
GPT-4: 34.9 (#268), gpt-oss-120b: 46.5 (#217)
| Benchmark | GPT-4 | gpt-oss-120b |
|---|---|---|
| LMArena Text | 1263 | 1365 |
| LMArena Creative Writing | 1244 | 1275 |
| EQ-Bench Creative Writing | 752 | 961 |
| LMArena Multi-Turn | 1257 | 1340 |
| Short-Story Creative Writing | — | 77.1% |
| WildBench | — | 84.5% |
Frequently asked questions
Is GPT-4 better than gpt-oss-120b?
gpt-oss-120b is the stronger model overall, scoring 36.3 to 29.1 on the Noometry Index.
Which is cheaper, GPT-4 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-4 lists at $30 and $60.
Is GPT-4 or gpt-oss-120b better for coding?
gpt-oss-120b scores higher on coding benchmarks: 33.5 versus 31.6 in the Noometry coding category.
Which has the bigger context window?
gpt-oss-120b does, with 131K tokens against 8K.
How many benchmarks do GPT-4 and gpt-oss-120b share?
27 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and gpt-oss-120b has 48.