Model comparison
gpt-oss-20b vs Qwen3 32B
Qwen3 32B is the stronger model overall, scoring 38.8 to 32.1 on the Noometry Index. gpt-oss-20b costs 34× less per token, which makes it the better buy when Qwen3 32B's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
Alibaba (Qwen)Summary
- They share 22 benchmarks with published results for both. gpt-oss-20b scores higher in 0 categories and Qwen3 32B in 9 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Qwen3 32B leads 32.3 to 9.2.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $0.70 / $2.80 for Qwen3 32B.
Side by side
| gpt-oss-20b | Qwen3 32B | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 32.1 | 38.8 |
| Released | 2025-08-05 | 2025-04 |
| Weights | Open | Open |
| Context window | 131K | 131K |
| Max output | 16K | 16K |
| Input $ / M tokens | $0.018 | $0.70 |
| Output $ / M tokens | $0.09 | $2.80 |
| Results tracked | 34 | 26 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
gpt-oss-20b: 37.1 (#185), Qwen3 32B: 37.3 (#184)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| SciCode | 34.4% | 35.4% |
| LMArena Coding | 1307 | 1358 |
| Aider Polyglot | — | 40% |
| WeirdML | 40.9% | — |
| ALE-Bench | 566.05 | — |
Agentic & Tool Use Qwen3 32B leads
gpt-oss-20b: 9.2 (#146), Qwen3 32B: 32.3 (#58)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| Terminal-Bench | 3.4% | — |
| Berkeley Function Calling Leaderboard | — | 48.7% |
Reasoning Too close to call
gpt-oss-20b: 18.8 (#251), Qwen3 32B: 19.6 (#233)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| Kagi LLM Benchmark | 53.2% | 54.9% |
| CritPt | 1.4% | 0.3% |
| Chess Puzzles | 4% | 5% |
| LMArena Hard Prompts | 1274 | 1335 |
| DTBench | 68% | 67.5% |
| LMCA | 14.5% | 17.3% |
| Epoch Capabilities Index | 137.81 | 138.51 |
Math Too close to call
gpt-oss-20b: 37.6 (#93), Qwen3 32B: 37.9 (#89)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 65.3% | 66.9% |
| LMArena Math | 1317 | 1399 |
| Omni-MATH | 56.5% | — |
Knowledge Qwen3 32B leads
gpt-oss-20b: 34.6 (#186), Qwen3 32B: 40.0 (#112)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| GPQA Diamond | 60.8% | 65.7% |
| LMArena Expert | 1258 | 1362 |
| MMLU-Pro | 74% | — |
| Vectara Hallucination Rate | — | 5.9% |
| GPQA (HELM) | 59.4% | — |
Multilingual Qwen3 32B leads
gpt-oss-20b: 42.4 (#192), Qwen3 32B: 45.8 (#162)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| LMArena Non-English | 1269 | 1317 |
| LMArena Chinese | 1311 | 1357 |
| LMArena German | 1249 | 1341 |
| LMArena Russian | 1279 | 1311 |
| LMArena Japanese | 1245 | — |
| LMArena Korean | 1239 | — |
| LMArena Spanish | 1267 | — |
Instruction Following Qwen3 32B leads
gpt-oss-20b: 62.0 (#235), Qwen3 32B: 69.1 (#174)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| LMArena Instruction Following | 1236 | 1305 |
| IFEval | 73.2% | — |
Long Context Qwen3 32B leads
gpt-oss-20b: 37.9 (#203), Qwen3 32B: 43.7 (#83)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| LMArena Longer Query | 1251 | 1327 |
| Fiction.LiveBench | — | 74.2% |
Writing & Preference Qwen3 32B leads
gpt-oss-20b: 35.8 (#260), Qwen3 32B: 53.1 (#158)
| Benchmark | gpt-oss-20b | Qwen3 32B |
|---|---|---|
| LMArena Text | 1288 | 1340 |
| LMArena Creative Writing | 1201 | 1297 |
| LMArena Multi-Turn | 1268 | 1331 |
| EQ-Bench Creative Writing | 666 | — |
| WildBench | 73.7% | — |
Frequently asked questions
Is gpt-oss-20b better than Qwen3 32B?
Qwen3 32B is the stronger model overall, scoring 38.8 to 32.1 on the Noometry Index. gpt-oss-20b costs 34× less per token, which makes it the better buy when Qwen3 32B's lead doesn't matter for your workload.
Which is cheaper, gpt-oss-20b or Qwen3 32B?
gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; Qwen3 32B lists at $0.70 and $2.80.
Is gpt-oss-20b or Qwen3 32B better for coding?
They score almost the same on coding (37.1 vs 37.3); test both on your own repository before choosing.
Which has the bigger context window?
Both accept 131K tokens.
How many benchmarks do gpt-oss-20b and Qwen3 32B share?
22 benchmarks have published results for both models. gpt-oss-20b has 34 scored results on Noometry and Qwen3 32B has 26.