Model comparison
GPT-5.4 Pro vs QwQ-32B
GPT-5.4 Pro is the stronger model overall, scoring 58.9 to 39.8 on the Noometry Index.
Last verified . 3 shared benchmarks.
Alibaba (Qwen)Summary
- They share 3 benchmarks with published results for both. GPT-5.4 Pro scores higher in 4 categories and QwQ-32B in 0 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 Pro leads 70.7 to 23.7.
- The biggest single-benchmark swing is Chess Puzzles: 58.6% for GPT-5.4 Pro and 5% for QwQ-32B.
- QwQ-32B has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 Pro | QwQ-32B | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 58.9 | 39.8 |
| Released | 2026-03-05 | 2024-11-28 |
| Weights | Proprietary | Open |
| Context window | 1.05M | — |
| Max output | 128K | — |
| Input $ / M tokens | $30 | — |
| Output $ / M tokens | $180 | — |
| Results tracked | 16 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 Pro leads
GPT-5.4 Pro: 43.2 (#90), QwQ-32B: 35.4 (#226)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| Aider Polyglot | — | 20.9% |
| WeirdML | 57.4% | — |
| BigCodeBench Instruct | — | 44.6% |
| LiveBench Coding | — | 72.2% |
| LMArena Coding | — | 1333 |
| BigCodeBench Complete | — | 54.4% |
Reasoning GPT-5.4 Pro leads
GPT-5.4 Pro: 70.7 (#13), QwQ-32B: 23.7 (#174)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| Chess Puzzles | 58.6% | 5% |
| Epoch Capabilities Index | 158.93 | 137.6 |
| ARC-AGI-2 | 83.3% | — |
| SimpleBench | 74.1% | — |
| ARC-AGI-1 | 94.5% | — |
| CritPt | 30% | — |
| EnigmaEval | 23.8% | — |
| LiveBench Reasoning | — | 83.5% |
| LMArena Hard Prompts | — | 1325 |
| LiveBench Data Analysis | — | 65% |
| ForecastBench | — | 58.3 |
| LiveBench | — | 72% |
Math GPT-5.4 Pro leads
GPT-5.4 Pro: 72.4 (#22), QwQ-32B: 38.0 (#143)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| FrontierMath (Tiers 1-3) | 82.5% | — |
| FrontierMath Tier 4 | 58.5% | — |
| OTIS Mock AIME 2024-2025 | — | 59.2% |
| LiveBench Math | — | 77.8% |
| LMArena Math | — | 1359 |
| FrontierMath (Feb 2025 set) | 50% | — |
| FrontierMath Tier 4 (v1) | 37.5% | — |
Knowledge GPT-5.4 Pro leads
GPT-5.4 Pro: 68.3 (#7), QwQ-32B: 37.2 (#158)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| GPQA Diamond | 94.6% | 65.3% |
| Humanity's Last Exam | 44.3% | — |
| SimpleQA Verified | 46.3% | — |
| Confabulations | — | 15.6% |
| Vectara Hallucination Rate | 8.3% | — |
| LMArena Expert | — | 1324 |
Multilingual Not comparable
GPT-5.4 Pro: —, QwQ-32B: 44.8 (#176)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| LMArena Non-English | — | 1305 |
| LMArena Chinese | — | 1378 |
| LMArena French | — | 1336 |
| LMArena German | — | 1313 |
| LMArena Japanese | — | 1262 |
| LMArena Korean | — | 1279 |
| LMArena Russian | — | 1297 |
| LMArena Spanish | — | 1354 |
Instruction Following Not comparable
GPT-5.4 Pro: —, QwQ-32B: 72.6 (#137)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| LiveBench Instruction Following | — | 81.8% |
| LMArena Instruction Following | — | 1297 |
Long Context Not comparable
GPT-5.4 Pro: —, QwQ-32B: 49.0 (#11)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| Fiction.LiveBench | — | 83.3% |
| LMArena Longer Query | — | 1308 |
Writing & Preference Not comparable
GPT-5.4 Pro: —, QwQ-32B: 50.6 (#180)
| Benchmark | GPT-5.4 Pro | QwQ-32B |
|---|---|---|
| LMArena Text | — | 1329 |
| LMArena Creative Writing | — | 1288 |
| Short-Story Creative Writing | — | 80.2% |
| EQ-Bench Creative Writing | — | 1257 |
| LMArena Multi-Turn | — | 1314 |
| LiveBench Language | — | 51.4% |
Frequently asked questions
Is GPT-5.4 Pro better than QwQ-32B?
GPT-5.4 Pro is the stronger model overall, scoring 58.9 to 39.8 on the Noometry Index.
Is GPT-5.4 Pro or QwQ-32B better for coding?
GPT-5.4 Pro scores higher on coding benchmarks: 43.2 versus 35.4 in the Noometry coding category.
How many benchmarks do GPT-5.4 Pro and QwQ-32B share?
3 benchmarks have published results for both models. GPT-5.4 Pro has 16 scored results on Noometry and QwQ-32B has 36.