# gpt-oss-120b vs o4-mini

> o4-mini is the stronger model overall, scoring 41.6 to 36.0 on the Noometry Index. gpt-oss-120b costs 27× less per token, which makes it the better buy when o4-mini's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-oss-120b-vs-o4-mini
- Last updated: 2026-10-10
- Shared benchmarks: 42

## Summary

- They share 42 benchmarks with published results for both. gpt-oss-120b scores higher in 2 categories and o4-mini in 7 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where o4-mini leads 32.3 to 12.1.
- The biggest single-benchmark swing is Fiction.LiveBench: 44.4% for gpt-oss-120b and 77.8% for o4-mini.
- gpt-oss-120b is cheaper at $0.037 / $0.17 per million input/output tokens, against $1.10 / $4.40 for o4-mini.
- o4-mini accepts more context: 200K tokens versus 131K.
- gpt-oss-120b has downloadable open weights; the other is API-only.

## Snapshot

| | gpt-oss-120b | o4-mini |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 36.0 | 41.6 |
| Rank | 210 | 119 |
| Context | 131K | 200K |
| Input $/M | $0.037 | $1.10 |
| Output $/M | $0.17 | $4.40 |
| Weights | Open | Proprietary |

## Coding

- gpt-oss-120b: 33.2 (#251)
- o4-mini: 41.6 (#110)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| SWE-bench Verified (bash only) | 26% | 45% |
| Aider Polyglot | 41.8% | 72% |
| WeirdML | 48.2% | 52.6% |
| LMArena Coding | 1380 | 1369 |
| ALE-Bench | 575.62 | 826.17 |
| AlgoTune | 1.41 | 1.72 |
| SciCode | 36% | — |
| GSO | — | 3.6% |
| CadEval | — | 62% |

## Agentic & Tool Use

- gpt-oss-120b: 12.1 (#145)
- o4-mini: 32.3 (#57)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| METR Time Horizons | 56.6% | 63.9% |
| Terminal-Bench | 18.7% | — |
| APEX-Agents | 4.4% | — |
| Berkeley Function Calling Leaderboard | — | 53.2% |
| GDPval | — | 25.3% |
| Vending-Bench 2 | -21.53 | — |

## Reasoning

- gpt-oss-120b: 19.5 (#236)
- o4-mini: 24.6 (#143)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| SimpleBench | 22.1% | 38.7% |
| Kagi LLM Benchmark | 58.6% | 67.6% |
| CritPt | 1.1% | 0.6% |
| Chess Puzzles | 20% | 26% |
| LMArena Hard Prompts | 1364 | 1351 |
| Mystery Game Puzzles | 2% | 5% |
| DTBench | 76.3% | 77.6% |
| LMCA | 22.1% | 26.5% |
| Epoch Capabilities Index | 139.94 | 145.63 |
| ARC-AGI-2 | — | 6.1% |
| ARC-AGI-1 | — | 58.7% |
| EnigmaEval | — | 9.2% |
| Surface Evolver Bench | 25% | — |
| ForecastBench | — | 61.8 |

## Math

- gpt-oss-120b: 50.6 (#43)
- o4-mini: 40.2 (#79)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9% | 81.7% |
| Omni-MATH | 68.8% | 72% |
| LMArena Math | 1388 | 1390 |
| FrontierMath (Tiers 1-3) | — | 36.1% |
| FrontierMath Tier 4 | — | 4.9% |
| MATH Level 5 | — | 97.8% |
| FrontierMath (Feb 2025 set) | — | 24.8% |
| FrontierMath Tier 4 (v1) | — | 6.3% |

## Knowledge

- gpt-oss-120b: 42.3 (#89)
- o4-mini: 43.5 (#83)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| GPQA Diamond | 75.8% | 79.6% |
| MMLU-Pro | 79.5% | 82% |
| Confabulations | 15.7% | 15.8% |
| Vectara Hallucination Rate | 14.2% | 18.6% |
| GPQA (HELM) | 68.4% | 73.5% |
| LMArena Expert | 1356 | 1344 |
| Humanity's Last Exam | — | 18.1% |
| SimpleQA Verified | — | 19.6% |

## Multimodal

- gpt-oss-120b: —
- o4-mini: 39.5 (#46)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| LMArena Vision | — | 1194 |
| GeoBench | — | 64% |
| VPCT | — | 57.5% |

## Multilingual

- gpt-oss-120b: 48.2 (#142)
- o4-mini: 47.2 (#149)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| LMArena Non-English | 1350 | 1336 |
| LMArena Chinese | 1384 | 1355 |
| LMArena French | 1370 | 1362 |
| LMArena German | 1355 | 1336 |
| LMArena Japanese | 1332 | 1308 |
| LMArena Korean | 1283 | 1313 |
| LMArena Russian | 1343 | 1334 |
| LMArena Spanish | 1388 | 1346 |

## Instruction Following

- gpt-oss-120b: 69.5 (#168)
- o4-mini: 75.4 (#63)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| IFEval | 83.6% | 92.8% |
| LMArena Instruction Following | 1318 | 1321 |

## Long Context

- gpt-oss-120b: 31.4 (#273)
- o4-mini: 45.5 (#31)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| Fiction.LiveBench | 44.4% | 77.8% |
| LMArena Longer Query | 1319 | 1315 |

## Writing & Preference

- gpt-oss-120b: 46.7 (#213)
- o4-mini: 54.2 (#148)

| Benchmark | gpt-oss-120b | o4-mini |
|---|---|---|
| LMArena Text | 1366 | 1353 |
| LMArena Creative Writing | 1274 | 1294 |
| Short-Story Creative Writing | 77.1% | 75% |
| WildBench | 84.5% | 85.4% |
| LMArena Multi-Turn | 1340 | 1350 |
| EQ-Bench Creative Writing | 961 | — |

## FAQ

### Is gpt-oss-120b better than o4-mini?

o4-mini is the stronger model overall, scoring 41.6 to 36.0 on the Noometry Index. gpt-oss-120b costs 27× less per token, which makes it the better buy when o4-mini's lead doesn't matter for your workload.

### Which is cheaper, gpt-oss-120b or o4-mini?

gpt-oss-120b is cheaper. It lists at $0.037 per million input tokens and $0.17 per million output tokens; o4-mini lists at $1.10 and $4.40.

### Is gpt-oss-120b or o4-mini better for coding?

o4-mini scores higher on coding benchmarks: 41.6 versus 33.2 in the Noometry coding category.

### Which has the bigger context window?

o4-mini does, with 200K tokens against 131K.

### How many benchmarks do gpt-oss-120b and o4-mini share?

42 benchmarks have published results for both models. gpt-oss-120b has 48 scored results on Noometry and o4-mini has 60.