Model comparison
gpt-oss-20b vs Llama 4 Maverick
gpt-oss-20b is the stronger model overall, scoring 32.1 to 30.6 on the Noometry Index.
Last verified . 32 shared benchmarks.
MetaSummary
- They share 32 benchmarks with published results for both. gpt-oss-20b scores higher in 5 categories and Llama 4 Maverick in 4 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Llama 4 Maverick leads 27.9 to 9.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 65.3% for gpt-oss-20b and 20.6% for Llama 4 Maverick.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $0.19 / $0.65 for Llama 4 Maverick.
- gpt-oss-20b accepts more context: 131K tokens versus 128K.
Side by side
| gpt-oss-20b | Llama 4 Maverick | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 32.1 | 30.6 |
| Released | 2025-08-05 | 2025-04-05 |
| Weights | Open | Open |
| Context window | 131K | 128K |
| Max output | 16K | 4K |
| Input $ / M tokens | $0.018 | $0.19 |
| Output $ / M tokens | $0.09 | $0.65 |
| Results tracked | 34 | 54 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding gpt-oss-20b leads
gpt-oss-20b: 37.1 (#185), Llama 4 Maverick: 26.3 (#318)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| SciCode | 34.4% | 33.1% |
| WeirdML | 40.9% | 24.5% |
| LMArena Coding | 1307 | 1302 |
| ALE-Bench | 566.05 | 172.97 |
| SWE-bench Verified (bash only) | — | 21% |
| Aider Polyglot | — | 15.6% |
| BigCodeBench Instruct | — | 49.7% |
| BigCodeBench Complete | — | 61.4% |
Agentic & Tool Use Llama 4 Maverick leads
gpt-oss-20b: 9.2 (#146), Llama 4 Maverick: 27.9 (#86)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| Terminal-Bench | 3.4% | — |
| Berkeley Function Calling Leaderboard | — | 37.3% |
Reasoning gpt-oss-20b leads
gpt-oss-20b: 18.8 (#251), Llama 4 Maverick: 10.1 (#337)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| Kagi LLM Benchmark | 53.2% | 55.9% |
| CritPt | 1.4% | 0% |
| LMArena Hard Prompts | 1274 | 1281 |
| DTBench | 68% | 61.9% |
| LMCA | 14.5% | 15.9% |
| Epoch Capabilities Index | 137.81 | 132.2 |
| ARC-AGI-2 | — | 0% |
| SimpleBench | — | 27.7% |
| NYT Connections (extended) | — | 8% |
| ARC-AGI-1 | — | 4.4% |
| Chess Puzzles | 4% | — |
| EnigmaEval | — | 0.6% |
| ForecastBench | — | 57.5 |
Math gpt-oss-20b leads
gpt-oss-20b: 37.6 (#93), Llama 4 Maverick: 24.6 (#256)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 65.3% | 20.6% |
| Omni-MATH | 56.5% | 42.2% |
| LMArena Math | 1317 | 1299 |
| MATH Level 5 | — | 73% |
| FrontierMath (Feb 2025 set) | — | 0.7% |
Knowledge gpt-oss-20b leads
gpt-oss-20b: 34.6 (#186), Llama 4 Maverick: 33.3 (#196)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| GPQA Diamond | 60.8% | 67% |
| MMLU-Pro | 74% | 81% |
| GPQA (HELM) | 59.4% | 65% |
| LMArena Expert | 1258 | 1259 |
| Humanity's Last Exam | — | 5.7% |
| Confabulations | — | 22.6% |
| Vectara Hallucination Rate | — | 8.2% |
Multimodal Not comparable
gpt-oss-20b: —, Llama 4 Maverick: 30.9 (#100)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| LMArena Vision | — | 1141 |
| GeoBench | — | 52% |
| SpatialViz-Bench | — | 31.8% |
Multilingual Too close to call
gpt-oss-20b: 42.4 (#192), Llama 4 Maverick: 42.4 (#190)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| LMArena Non-English | 1269 | 1269 |
| LMArena Chinese | 1311 | 1277 |
| LMArena German | 1249 | 1291 |
| LMArena Japanese | 1245 | 1207 |
| LMArena Korean | 1239 | 1203 |
| LMArena Russian | 1279 | 1286 |
| LMArena Spanish | 1267 | 1293 |
| LMArena French | — | 1258 |
Instruction Following Llama 4 Maverick leads
gpt-oss-20b: 62.0 (#235), Llama 4 Maverick: 71.9 (#140)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| IFEval | 73.2% | 90.8% |
| LMArena Instruction Following | 1236 | 1267 |
Long Context gpt-oss-20b leads
gpt-oss-20b: 37.9 (#203), Llama 4 Maverick: 31.4 (#274)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| LMArena Longer Query | 1251 | 1280 |
| Fiction.LiveBench | — | 46.2% |
Writing & Preference Llama 4 Maverick leads
gpt-oss-20b: 35.8 (#260), Llama 4 Maverick: 39.0 (#247)
| Benchmark | gpt-oss-20b | Llama 4 Maverick |
|---|---|---|
| LMArena Text | 1288 | 1287 |
| LMArena Creative Writing | 1201 | 1266 |
| EQ-Bench Creative Writing | 666 | 860 |
| WildBench | 73.7% | 80% |
| LMArena Multi-Turn | 1268 | 1289 |
| Short-Story Creative Writing | — | 62% |
Frequently asked questions
Is gpt-oss-20b better than Llama 4 Maverick?
gpt-oss-20b is the stronger model overall, scoring 32.1 to 30.6 on the Noometry Index.
Which is cheaper, gpt-oss-20b or Llama 4 Maverick?
gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; Llama 4 Maverick lists at $0.19 and $0.65.
Is gpt-oss-20b or Llama 4 Maverick better for coding?
gpt-oss-20b scores higher on coding benchmarks: 37.1 versus 26.3 in the Noometry coding category.
Which has the bigger context window?
gpt-oss-20b does, with 131K tokens against 128K.
How many benchmarks do gpt-oss-20b and Llama 4 Maverick share?
32 benchmarks have published results for both models. gpt-oss-20b has 34 scored results on Noometry and Llama 4 Maverick has 54.