# gpt-oss-120b vs Llama-3.3-70B-Instruct

> gpt-oss-120b is the stronger model overall, scoring 36.3 to 30.6 on the Noometry Index.

- Canonical page: https://noometry.com/compare/gpt-oss-120b-vs-llama-3-3-70b-instruct
- Last updated: 2026-10-12
- Shared benchmarks: 29

## Summary

- They share 29 benchmarks with published results for both. gpt-oss-120b scores higher in 6 categories and Llama-3.3-70B-Instruct in 3 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where gpt-oss-120b leads 52.5 to 15.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 88.9% for gpt-oss-120b and 5.1% for Llama-3.3-70B-Instruct.
- gpt-oss-120b is cheaper at $0.037 / $0.17 per million input/output tokens, against $0.10 / $0.32 for Llama-3.3-70B-Instruct.
- gpt-oss-120b accepts more context: 131K tokens versus 128K.

## Snapshot

| | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 36.3 | 30.6 |
| Rank | 217 | 291 |
| Context | 131K | 128K |
| Input $/M | $0.037 | $0.10 |
| Output $/M | $0.17 | $0.32 |
| Weights | Open | Open |

## Coding

- gpt-oss-120b: 33.5 (#256)
- Llama-3.3-70B-Instruct: 31.0 (#290)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| SciCode | 36% | 26% |
| WeirdML | 48.2% | 14.4% |
| LMArena Coding | 1380 | 1268 |
| SWE-bench Verified (bash only) | 26% | — |
| Aider Polyglot | 41.8% | — |
| BigCodeBench Instruct | — | 46.9% |
| LiveBench Coding | — | 36.6% |
| BigCodeBench Complete | — | 57.5% |
| ALE-Bench | 575.62 | — |
| AlgoTune | 1.41 | — |

## Agentic & Tool Use

- gpt-oss-120b: 12.2 (#153)
- Llama-3.3-70B-Instruct: 25.8 (#105)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| Terminal-Bench | 18.7% | — |
| APEX-Agents | 4.4% | — |
| Berkeley Function Calling Leaderboard | — | 31.9% |
| BALROG | — | 23% |
| METR Time Horizons | 56.6% | — |
| Vending-Bench 2 | -21.53 | — |

## Reasoning

- gpt-oss-120b: 20.0 (#245)
- Llama-3.3-70B-Instruct: 14.1 (#327)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| SimpleBench | 22.1% | 19.9% |
| CritPt | 1.1% | 0% |
| LMArena Hard Prompts | 1364 | 1257 |
| DTBench | 76.3% | 59.5% |
| LMCA | 22.1% | 17.5% |
| Epoch Capabilities Index | 139.93 | 127.33 |
| Kagi LLM Benchmark | 58.6% | — |
| Chess Puzzles | 20% | — |
| LiveBench Reasoning | — | 50.8% |
| Mystery Game Puzzles | 2% | — |
| LiveBench Data Analysis | — | 49.5% |
| Surface Evolver Bench | 25% | — |
| ForecastBench | — | 58.6 |
| LiveBench | — | 50.2% |

## Math

- gpt-oss-120b: 52.5 (#50)
- Llama-3.3-70B-Instruct: 15.3 (#298)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9% | 5.1% |
| LMArena Math | 1389 | 1267 |
| Omni-MATH | 68.8% | — |
| LiveBench Math | — | 42.2% |
| MATH Level 5 | — | 41.6% |

## Knowledge

- gpt-oss-120b: 42.4 (#96)
- Llama-3.3-70B-Instruct: 30.6 (#226)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| GPQA Diamond | 75.8% | 47.4% |
| Confabulations | 15.7% | 22.8% |
| Vectara Hallucination Rate | 14.2% | 4.1% |
| LMArena Expert | 1356 | 1225 |
| MMLU-Pro | 79.5% | — |
| GPQA (HELM) | 68.4% | — |
| MMLU | — | 86.3% |

## Multilingual

- gpt-oss-120b: 48.0 (#147)
- Llama-3.3-70B-Instruct: 39.9 (#220)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1351 | 1236 |
| LMArena Chinese | 1385 | 1217 |
| LMArena French | 1369 | 1281 |
| LMArena German | 1353 | 1251 |
| LMArena Japanese | 1331 | 1150 |
| LMArena Korean | 1282 | 1143 |
| LMArena Russian | 1343 | 1252 |
| LMArena Spanish | 1389 | 1270 |

## Instruction Following

- gpt-oss-120b: 69.3 (#173)
- Llama-3.3-70B-Instruct: 71.1 (#157)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Instruction Following | 1318 | 1242 |
| LiveBench Instruction Following | — | 82.7% |
| IFEval | 83.6% | — |

## Long Context

- gpt-oss-120b: 31.4 (#278)
- Llama-3.3-70B-Instruct: 26.4 (#295)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| Fiction.LiveBench | 44.4% | 33.3% |
| LMArena Longer Query | 1319 | 1256 |

## Writing & Preference

- gpt-oss-120b: 46.5 (#217)
- Llama-3.3-70B-Instruct: 47.6 (#207)

| Benchmark | gpt-oss-120b | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1365 | 1274 |
| LMArena Creative Writing | 1275 | 1250 |
| LMArena Multi-Turn | 1340 | 1280 |
| Short-Story Creative Writing | 77.1% | — |
| EQ-Bench Creative Writing | 961 | — |
| WildBench | 84.5% | — |
| LiveBench Language | — | 39.2% |

## FAQ

### Is gpt-oss-120b better than Llama-3.3-70B-Instruct?

gpt-oss-120b is the stronger model overall, scoring 36.3 to 30.6 on the Noometry Index.

### Which is cheaper, gpt-oss-120b or Llama-3.3-70B-Instruct?

gpt-oss-120b is cheaper. It lists at $0.037 per million input tokens and $0.17 per million output tokens; Llama-3.3-70B-Instruct lists at $0.10 and $0.32.

### Is gpt-oss-120b or Llama-3.3-70B-Instruct better for coding?

gpt-oss-120b scores higher on coding benchmarks: 33.5 versus 31.0 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-oss-120b and Llama-3.3-70B-Instruct share?

29 benchmarks have published results for both models. gpt-oss-120b has 48 scored results on Noometry and Llama-3.3-70B-Instruct has 43.