Model comparison
GPT-5.2 vs Qwen3 32B
GPT-5.2 is the stronger model overall, scoring 54.1 to 39.2 on the Noometry Index. Qwen3 32B costs 3.9× less per token, which makes it the better buy when GPT-5.2'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-5.2 scores higher in 9 categories and Qwen3 32B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 20.2.
- The biggest single-benchmark swing is Chess Puzzles: 49% for GPT-5.2 and 5% for Qwen3 32B.
- Qwen3 32B is cheaper at $0.70 / $2.80 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 131K.
- Qwen3 32B has downloadable open weights; the other is API-only.
Side by side
| GPT-5.2 | Qwen3 32B | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 54.1 | 39.2 |
| Released | 2025-12-11 | 2025-04 |
| Weights | Proprietary | Open |
| Context window | 400K | 131K |
| Max output | 128K | 16K |
| Input $ / M tokens | $1.75 | $0.70 |
| Output $ / M tokens | $14 | $2.80 |
| Results tracked | 67 | 26 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GPT-5.2: 51.6 (#37), Qwen3 32B: 37.7 (#190)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| LMArena Coding | 1447 | 1358 |
| SWE-bench Verified | 73.8% | — |
| SWE-bench Verified (bash only) | 72.8% | — |
| Aider Polyglot | — | 40% |
| LMArena WebDev | 1416 | — |
| SWE-bench Multilingual | 66.7% | — |
| SciCode | — | 35.4% |
| GSO | 27.4% | — |
| WeirdML | 72.2% | — |
| ALE-Bench | 1,294 | — |
| AlgoTune | 2.05 | — |
Agentic & Tool Use GPT-5.2 leads
GPT-5.2: 40.2 (#24), Qwen3 32B: 32.6 (#62)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 55.9% | 48.7% |
| Terminal-Bench | 64.9% | — |
| GDPval | 49.7% | — |
| Remote Labor Index | 2.5% | — |
| τ²-bench Airline | 83% | — |
| τ²-bench Banking | 32.2% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| DeepResearch Bench | 41.1% | — |
| LMArena Search | 1207 | — |
| METR Time Horizons | 75.3% | — |
| Vending-Bench 2 | 3,591 | — |
Reasoning GPT-5.2 leads
GPT-5.2: 50.2 (#35), Qwen3 32B: 20.2 (#241)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| Kagi LLM Benchmark | 73.3% | 54.9% |
| Chess Puzzles | 49% | 5% |
| LMArena Hard Prompts | 1445 | 1334 |
| DTBench | 90.9% | 67.5% |
| LMCA | 43.9% | 17.3% |
| Epoch Capabilities Index | 153.45 | 138.51 |
| ARC-AGI-2 | 52.9% | — |
| SimpleBench | 45.8% | — |
| NYT Connections (extended) | 83.6% | — |
| ARC-AGI-1 | 86.2% | — |
| CritPt | — | 0.3% |
| EnigmaEval | 10.4% | — |
| EBR-Bench | 23% | — |
| Mystery Game Puzzles | 23% | — |
| ForecastBench | 60.1 | — |
Math GPT-5.2 leads
GPT-5.2: 60.0 (#38), Qwen3 32B: 39.7 (#99)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 96.1% | 66.9% |
| LMArena Math | 1440 | 1399 |
| FrontierMath (Tiers 1-3) | 67.4% | — |
| FrontierMath Tier 4 | 31.7% | — |
| MathArena Final-Answer Competitions | 72% | — |
| ProofBench | 15% | — |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Tier 4 (v1) | 18.8% | — |
Knowledge GPT-5.2 leads
GPT-5.2: 59.3 (#32), Qwen3 32B: 40.0 (#125)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| GPQA Diamond | 91.4% | 65.7% |
| Vectara Hallucination Rate | 8.4% | 5.9% |
| LMArena Expert | 1445 | 1362 |
| Humanity's Last Exam | 27.8% | — |
| SimpleQA Verified | 37.1% | — |
Multimodal Not comparable
GPT-5.2: 51.3 (#7), Qwen3 32B: —
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| LMArena Vision | 1268 | — |
| VPCT | 84% | — |
| Furniture Assembly | 38.3% | — |
| LMArena Document | 1405 | — |
Multilingual GPT-5.2 leads
GPT-5.2: 53.4 (#67), Qwen3 32B: 45.6 (#167)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| LMArena Non-English | 1425 | 1317 |
| LMArena Chinese | 1460 | 1357 |
| LMArena German | 1448 | 1341 |
| LMArena Russian | 1440 | 1311 |
| LMArena French | 1455 | — |
| LMArena Japanese | 1420 | — |
| LMArena Korean | 1392 | — |
| LMArena Spanish | 1433 | — |
Instruction Following GPT-5.2 leads
GPT-5.2: 74.7 (#89), Qwen3 32B: 68.9 (#179)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| LMArena Instruction Following | 1417 | 1305 |
Long Context Too close to call
GPT-5.2: 44.0 (#78), Qwen3 32B: 43.8 (#87)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| LMArena Longer Query | 1428 | 1327 |
| Fiction.LiveBench | — | 74.2% |
| CL-bench | 18.2% | — |
Writing & Preference GPT-5.2 leads
GPT-5.2: 66.8 (#32), Qwen3 32B: 52.9 (#163)
| Benchmark | GPT-5.2 | Qwen3 32B |
|---|---|---|
| LMArena Text | 1439 | 1340 |
| LMArena Creative Writing | 1401 | 1297 |
| LMArena Multi-Turn | 1458 | 1331 |
| EQ-Bench Creative Writing | 1703 | — |
Frequently asked questions
Is GPT-5.2 better than Qwen3 32B?
GPT-5.2 is the stronger model overall, scoring 54.1 to 39.2 on the Noometry Index. Qwen3 32B costs 3.9× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Which is cheaper, GPT-5.2 or Qwen3 32B?
Qwen3 32B is cheaper. It lists at $0.70 per million input tokens and $2.80 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GPT-5.2 or Qwen3 32B better for coding?
GPT-5.2 scores higher on coding benchmarks: 51.6 versus 37.7 in the Noometry coding category.
Which has the bigger context window?
GPT-5.2 does, with 400K tokens against 131K.
How many benchmarks do GPT-5.2 and Qwen3 32B share?
22 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and Qwen3 32B has 26.