Model comparison
gpt-oss-20b vs Llama 3-70B
gpt-oss-20b is the stronger model overall, scoring 32.5 to 28.8 on the Noometry Index.
Last verified . 21 shared benchmarks.
MetaSummary
- They share 21 benchmarks with published results for both. gpt-oss-20b scores higher in 6 categories and Llama 3-70B in 3 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where gpt-oss-20b leads 39.4 to 12.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 65.3% for gpt-oss-20b and 4.3% for Llama 3-70B.
Side by side
| gpt-oss-20b | Llama 3-70B | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 32.5 | 28.8 |
| Released | 2025-08-05 | 2024-04-18 |
| Weights | Open | Open |
| Context window | 131K | — |
| Max output | 16K | — |
| Input $ / M tokens | $0.018 | — |
| Output $ / M tokens | $0.09 | — |
| Results tracked | 34 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding gpt-oss-20b leads
gpt-oss-20b: 37.6 (#192), Llama 3-70B: 35.8 (#218)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| LMArena Coding | 1306 | 1206 |
| SciCode | 34.4% | — |
| WeirdML | 40.9% | — |
| BigCodeBench Instruct | — | 43.6% |
| BigCodeBench Complete | — | 54.5% |
| ALE-Bench | 566.05 | — |
| HumanEval+ | — | 72% |
| MBPP+ | — | 69% |
Agentic & Tool Use Llama 3-70B leads
gpt-oss-20b: 9.3 (#154), Llama 3-70B: 21.1 (#139)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| Terminal-Bench | 3.4% | — |
| Cybench | — | 5% |
Reasoning gpt-oss-20b leads
gpt-oss-20b: 19.3 (#261), Llama 3-70B: 18.0 (#288)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| Kagi LLM Benchmark | 53.2% | 35.1% |
| LMArena Hard Prompts | 1274 | 1195 |
| DTBench | 68% | 54.2% |
| Epoch Capabilities Index | 137.82 | 122.93 |
| CritPt | 1.4% | — |
| Chess Puzzles | 4% | — |
| LMCA | 14.5% | — |
| ForecastBench | — | 57.1 |
| WinoGrande | — | 83.5% |
Math gpt-oss-20b leads
gpt-oss-20b: 39.4 (#103), Llama 3-70B: 12.8 (#305)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 65.3% | 4.3% |
| LMArena Math | 1317 | 1218 |
| Omni-MATH | 56.5% | — |
| MATH Level 5 | — | 22.6% |
Knowledge gpt-oss-20b leads
gpt-oss-20b: 34.6 (#195), Llama 3-70B: 20.8 (#277)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| GPQA Diamond | 60.8% | 40.6% |
| LMArena Expert | 1258 | 1149 |
| MMLU-Pro | 74% | — |
| GPQA (HELM) | 59.4% | — |
| MMLU | — | 79.3% |
Multilingual gpt-oss-20b leads
gpt-oss-20b: 42.2 (#197), Llama 3-70B: 33.6 (#251)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| LMArena Non-English | 1268 | 1142 |
| LMArena Chinese | 1314 | 1114 |
| LMArena German | 1255 | 1169 |
| LMArena Japanese | 1244 | 1017 |
| LMArena Korean | 1236 | 1017 |
| LMArena Russian | 1278 | 1159 |
| LMArena Spanish | 1267 | 1241 |
| LMArena French | — | 1232 |
Instruction Following Too close to call
gpt-oss-20b: 61.8 (#240), Llama 3-70B: 62.5 (#238)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| LMArena Instruction Following | 1236 | 1194 |
| IFEval | 73.2% | — |
Long Context gpt-oss-20b leads
gpt-oss-20b: 37.9 (#209), Llama 3-70B: 35.6 (#240)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| LMArena Longer Query | 1250 | 1174 |
Writing & Preference Llama 3-70B leads
gpt-oss-20b: 35.5 (#265), Llama 3-70B: 42.8 (#231)
| Benchmark | gpt-oss-20b | Llama 3-70B |
|---|---|---|
| LMArena Text | 1287 | 1221 |
| LMArena Creative Writing | 1201 | 1210 |
| LMArena Multi-Turn | 1268 | 1223 |
| EQ-Bench Creative Writing | 666 | — |
| WildBench | 73.7% | — |
Frequently asked questions
Is gpt-oss-20b better than Llama 3-70B?
gpt-oss-20b is the stronger model overall, scoring 32.5 to 28.8 on the Noometry Index.
Is gpt-oss-20b or Llama 3-70B better for coding?
gpt-oss-20b scores higher on coding benchmarks: 37.6 versus 35.8 in the Noometry coding category.
How many benchmarks do gpt-oss-20b and Llama 3-70B share?
21 benchmarks have published results for both models. gpt-oss-20b has 34 scored results on Noometry and Llama 3-70B has 31.