Model comparison
gpt-oss-120b vs o3
o3 is the stronger model overall, scoring 47.5 to 36.3 on the Noometry Index. gpt-oss-120b costs 50× less per token, which makes it the better buy when o3's lead doesn't matter for your workload.
Last verified . 41 shared benchmarks.
Summary
- They share 41 benchmarks with published results for both. gpt-oss-120b scores higher in 1 category and o3 in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where o3 leads 34.5 to 12.2.
- The biggest single-benchmark swing is Fiction.LiveBench: 44.4% for gpt-oss-120b and 88.9% for o3.
- gpt-oss-120b is cheaper at $0.037 / $0.17 per million input/output tokens, against $2 / $8 for o3.
- o3 accepts more context: 200K tokens versus 131K.
- gpt-oss-120b has downloadable open weights; the other is API-only.
Side by side
| gpt-oss-120b | o3 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 36.3 | 47.5 |
| Released | 2025-08-05 | 2025-04-16 |
| Weights | Open | Proprietary |
| Context window | 131K | 200K |
| Max output | 41K | 100K |
| Input $ / M tokens | $0.037 | $2 |
| Output $ / M tokens | $0.17 | $8 |
| Results tracked | 48 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o3 leads
gpt-oss-120b: 33.5 (#256), o3: 46.8 (#64)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| SWE-bench Verified (bash only) | 26% | 58.4% |
| Aider Polyglot | 41.8% | 81.3% |
| WeirdML | 48.2% | 52.4% |
| LMArena Coding | 1380 | 1408 |
| ALE-Bench | 575.62 | 933.55 |
| SWE-bench Verified | — | 62.3% |
| SciCode | 36% | — |
| GSO | — | 8.8% |
| CadEval | — | 74% |
| AlgoTune | 1.41 | — |
Agentic & Tool Use o3 leads
gpt-oss-120b: 12.2 (#153), o3: 34.5 (#44)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| METR Time Horizons | 56.6% | 65.4% |
| Terminal-Bench | 18.7% | — |
| APEX-Agents | 4.4% | — |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 30.8% |
| DeepResearch Bench | — | 45.2% |
| OSWorld | — | 23% |
| LMArena Search | — | 1144 |
| Vending-Bench 2 | -21.53 | — |
Reasoning o3 leads
gpt-oss-120b: 20.0 (#245), o3: 32.0 (#78)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| SimpleBench | 22.1% | 53.1% |
| Kagi LLM Benchmark | 58.6% | 67.6% |
| CritPt | 1.1% | 1.4% |
| Chess Puzzles | 20% | 38% |
| LMArena Hard Prompts | 1364 | 1402 |
| Mystery Game Puzzles | 2% | 29% |
| DTBench | 76.3% | 84.8% |
| LMCA | 22.1% | 39.7% |
| Epoch Capabilities Index | 139.93 | 146.86 |
| ARC-AGI-2 | — | 6.5% |
| ARC-AGI-1 | — | 60.8% |
| EnigmaEval | — | 13.1% |
| Surface Evolver Bench | 25% | — |
| ForecastBench | — | 62.5 |
Math gpt-oss-120b leads
gpt-oss-120b: 52.5 (#50), o3: 50.2 (#58)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9% | 84.4% |
| Omni-MATH | 68.8% | 71.4% |
| LMArena Math | 1389 | 1426 |
| FrontierMath (Tiers 1-3) | — | 33.3% |
| MATH Level 5 | — | 97.8% |
| FrontierMath (Feb 2025 set) | — | 18.7% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge o3 leads
gpt-oss-120b: 42.4 (#96), o3: 54.6 (#52)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| GPQA Diamond | 75.8% | 81.8% |
| MMLU-Pro | 79.5% | 85.9% |
| Confabulations | 15.7% | 14.4% |
| GPQA (HELM) | 68.4% | 75.3% |
| LMArena Expert | 1356 | 1402 |
| Humanity's Last Exam | — | 20.3% |
| SimpleQA Verified | — | 49.4% |
| Vectara Hallucination Rate | 14.2% | — |
Multimodal Not comparable
gpt-oss-120b: —, o3: 41.4 (#36)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| LMArena Vision | — | 1214 |
| GeoBench | — | 74% |
| VPCT | — | 52% |
Multilingual o3 leads
gpt-oss-120b: 48.0 (#147), o3: 51.7 (#105)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| LMArena Non-English | 1351 | 1401 |
| LMArena Chinese | 1385 | 1437 |
| LMArena French | 1369 | 1430 |
| LMArena German | 1353 | 1420 |
| LMArena Japanese | 1331 | 1403 |
| LMArena Korean | 1282 | 1370 |
| LMArena Russian | 1343 | 1406 |
| LMArena Spanish | 1389 | 1395 |
Instruction Following o3 leads
gpt-oss-120b: 69.3 (#173), o3: 72.8 (#127)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| IFEval | 83.6% | 86.9% |
| LMArena Instruction Following | 1318 | 1368 |
Long Context o3 leads
gpt-oss-120b: 31.4 (#278), o3: 53.3 (#6)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| Fiction.LiveBench | 44.4% | 88.9% |
| LMArena Longer Query | 1319 | 1372 |
| CL-bench | — | 17.8% |
Writing & Preference o3 leads
gpt-oss-120b: 46.5 (#217), o3: 63.5 (#64)
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
| LMArena Text | 1365 | 1410 |
| LMArena Creative Writing | 1275 | 1359 |
| Short-Story Creative Writing | 77.1% | 83.9% |
| EQ-Bench Creative Writing | 961 | 1676 |
| WildBench | 84.5% | 86.1% |
| LMArena Multi-Turn | 1340 | 1405 |
Frequently asked questions
Is gpt-oss-120b better than o3?
o3 is the stronger model overall, scoring 47.5 to 36.3 on the Noometry Index. gpt-oss-120b costs 50× less per token, which makes it the better buy when o3's lead doesn't matter for your workload.
Which is cheaper, gpt-oss-120b or o3?
gpt-oss-120b is cheaper. It lists at $0.037 per million input tokens and $0.17 per million output tokens; o3 lists at $2 and $8.
Is gpt-oss-120b or o3 better for coding?
o3 scores higher on coding benchmarks: 46.8 versus 33.5 in the Noometry coding category.
Which has the bigger context window?
o3 does, with 200K tokens against 131K.
How many benchmarks do gpt-oss-120b and o3 share?
41 benchmarks have published results for both models. gpt-oss-120b has 48 scored results on Noometry and o3 has 63.