Model comparison
GPT-5 vs Qwen3 235B-A22B
GPT-5 is the stronger model overall, scoring 50.9 to 43.2 on the Noometry Index. Qwen3 235B-A22B costs 2.8× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Last verified . 47 shared benchmarks.
Alibaba (Qwen)Summary
- They share 47 benchmarks with published results for both. GPT-5 scores higher in 7 categories and Qwen3 235B-A22B in 2 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-5 leads 69.5 to 46.1.
- The biggest single-benchmark swing is ARC-AGI-1: 65.7% for GPT-5 and 11% for Qwen3 235B-A22B.
- Qwen3 235B-A22B is cheaper at $0.70 / $2.80 per million input/output tokens, against $1.25 / $10 for GPT-5.
- GPT-5 accepts more context: 400K tokens versus 131K.
- Qwen3 235B-A22B has downloadable open weights; the other is API-only.
Side by side
| GPT-5 | Qwen3 235B-A22B | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 50.9 | 43.2 |
| Released | 2025-08-07 | 2025-04 |
| Weights | Proprietary | Open |
| Context window | 400K | 131K |
| Max output | 128K | 16K |
| Input $ / M tokens | $1.25 | $0.70 |
| Output $ / M tokens | $10 | $2.80 |
| Results tracked | 69 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 leads
GPT-5: 50.7 (#39), Qwen3 235B-A22B: 43.8 (#72)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| Aider Polyglot | 88% | 59.6% |
| SciCode | 42.9% | 42.4% |
| WeirdML | 60.7% | 41% |
| LMArena Coding | 1436 | 1445 |
| SWE-bench Verified | 73.6% | — |
| SWE-bench Verified (bash only) | 65% | — |
| LMArena WebDev | 1417 | — |
| GSO | 6.9% | — |
| ALE-Bench | 1,162 | — |
| AlgoTune | 1.67 | — |
Agentic & Tool Use Too close to call
GPT-5: 33.0 (#53), Qwen3 235B-A22B: 33.7 (#48)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| Terminal-Bench | 49.6% | — |
| Berkeley Function Calling Leaderboard | — | 52.1% |
| GDPval | 34.8% | — |
| Remote Labor Index | 1.7% | — |
| DeepResearch Bench | 49.6% | — |
| BALROG | 32.8% | — |
| LMArena Search | 1133 | — |
| METR Time Horizons | 69.6% | — |
| Vending-Bench 2 | — | -11.34 |
Reasoning GPT-5 leads
GPT-5: 38.3 (#55), Qwen3 235B-A22B: 15.8 (#302)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| ARC-AGI-2 | 9.9% | 1.3% |
| SimpleBench | 56.7% | 31% |
| Kagi LLM Benchmark | 72.7% | 69.4% |
| ARC-AGI-1 | 65.7% | 11% |
| CritPt | 12.6% | 0% |
| Chess Puzzles | 37% | 12% |
| LMArena Hard Prompts | 1415 | 1433 |
| Mystery Game Puzzles | 23% | 9% |
| DTBench | 90.7% | 80.3% |
| LMCA | 40% | 29.3% |
| Epoch Capabilities Index | 150 | 143.87 |
| ForecastBench | 61.4 | 59.7 |
| EnigmaEval | 10.5% | — |
| EBR-Bench | 12.7% | — |
Math GPT-5 leads
GPT-5: 54.6 (#36), Qwen3 235B-A22B: 48.5 (#51)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 91.4% | 86.7% |
| Omni-MATH | 64.7% | 71.8% |
| LMArena Math | 1406 | 1431 |
| MATH Level 5 | 98.1% | 68.9% |
| FrontierMath (Feb 2025 set) | 32.4% | 8.5% |
| FrontierMath Tier 4 (v1) | 12.5% | 0% |
| FrontierMath (Tiers 1-3) | 55.4% | — |
| FrontierMath Tier 4 | 22% | — |
| ProofBench | 18% | — |
Knowledge GPT-5 leads
GPT-5: 56.5 (#37), Qwen3 235B-A22B: 49.5 (#67)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| GPQA Diamond | 86.2% | 80.1% |
| SimpleQA Verified | 50.1% | 40.4% |
| MMLU-Pro | 86.3% | 84.4% |
| Confabulations | 10.3% | 15.6% |
| Vectara Hallucination Rate | 14.7% | 9.3% |
| GPQA (HELM) | 79.2% | 72.7% |
| LMArena Expert | 1420 | 1460 |
| Humanity's Last Exam | 25.3% | — |
Multimodal Not comparable
GPT-5: 46.1 (#12), Qwen3 235B-A22B: —
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Vision | 1232 | — |
| GeoBench | 81% | — |
| VPCT | 66% | — |
Multilingual Too close to call
GPT-5: 51.5 (#105), Qwen3 235B-A22B: 52.5 (#84)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Non-English | 1397 | 1410 |
| LMArena Chinese | 1423 | 1481 |
| LMArena French | 1409 | 1445 |
| LMArena German | 1415 | 1432 |
| LMArena Japanese | 1411 | 1400 |
| LMArena Korean | 1361 | 1391 |
| LMArena Russian | 1406 | 1411 |
| LMArena Spanish | 1398 | 1429 |
Instruction Following GPT-5 leads
GPT-5: 74.0 (#107), Qwen3 235B-A22B: 72.8 (#132)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| IFEval | 87.5% | 83.5% |
| LMArena Instruction Following | 1388 | 1408 |
Long Context GPT-5 leads
GPT-5: 69.5 (#2), Qwen3 235B-A22B: 46.1 (#25)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| Fiction.LiveBench | 97.2% | 75% |
| LMArena Longer Query | 1399 | 1426 |
Writing & Preference GPT-5 leads
GPT-5: 63.5 (#63), Qwen3 235B-A22B: 59.8 (#103)
| Benchmark | GPT-5 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Text | 1406 | 1419 |
| LMArena Creative Writing | 1365 | 1384 |
| Short-Story Creative Writing | 86% | 83% |
| EQ-Bench Creative Writing | 1627 | 1366 |
| WildBench | 85.7% | 86.6% |
| LMArena Multi-Turn | 1426 | 1432 |
Frequently asked questions
Is GPT-5 better than Qwen3 235B-A22B?
GPT-5 is the stronger model overall, scoring 50.9 to 43.2 on the Noometry Index. Qwen3 235B-A22B costs 2.8× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Which is cheaper, GPT-5 or Qwen3 235B-A22B?
Qwen3 235B-A22B is cheaper. It lists at $0.70 per million input tokens and $2.80 per million output tokens; GPT-5 lists at $1.25 and $10.
Is GPT-5 or Qwen3 235B-A22B better for coding?
GPT-5 scores higher on coding benchmarks: 50.7 versus 43.8 in the Noometry coding category.
Which has the bigger context window?
GPT-5 does, with 400K tokens against 131K.
How many benchmarks do GPT-5 and Qwen3 235B-A22B share?
47 benchmarks have published results for both models. GPT-5 has 69 scored results on Noometry and Qwen3 235B-A22B has 49.