Model comparison
GPT-5.2 vs Qwen3.8 Max
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 54.1 on the Noometry Index.
Last verified . 33 shared benchmarks.
Alibaba (Qwen)Summary
- They share 33 benchmarks with published results for both. GPT-5.2 scores higher in 1 category and Qwen3.8 Max in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in multimodal, where GPT-5.2 leads 51.3 to 37.2.
- The biggest single-benchmark swing is ProofBench: 15% for GPT-5.2 and 58% for Qwen3.8 Max.
- Qwen3.8 Max is cheaper at $2 / $6 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- Qwen3.8 Max accepts more context: 1M tokens versus 400K.
Side by side
| GPT-5.2 | Qwen3.8 Max | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 54.1 | 56.8 |
| Released | 2025-12-11 | 2026-08-02 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 1M |
| Max output | 128K | 131K |
| Input $ / M tokens | $1.75 | $2 |
| Output $ / M tokens | $14 | $6 |
| Results tracked | 67 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 Max leads
GPT-5.2: 51.6 (#37), Qwen3.8 Max: 53.5 (#29)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| LMArena WebDev | 1416 | 1674 |
| LMArena Coding | 1447 | 1502 |
| SWE-bench Verified | 73.8% | — |
| DeepSWE | — | 57.5% |
| SWE-bench Verified (bash only) | 72.8% | — |
| SWE-bench Multilingual | 66.7% | — |
| FrontierSWE | — | 17.8% |
| SciCode | — | 53.2% |
| GSO | 27.4% | — |
| WeirdML | 72.2% | — |
| ALE-Bench | 1,294 | — |
| AlgoTune | 2.05 | — |
Agentic & Tool Use Qwen3.8 Max leads
GPT-5.2: 40.2 (#24), Qwen3.8 Max: 45.4 (#14)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| τ²-bench Banking | 32.2% | 55.1% |
| Terminal-Bench | 64.9% | — |
| APEX-Agents | — | 63.3% |
| Berkeley Function Calling Leaderboard | 55.9% | — |
| GDPval | 49.7% | — |
| Remote Labor Index | 2.5% | — |
| τ²-bench Airline | 83% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| DeepResearch Bench | 41.1% | — |
| GDP.pdf | — | 23.2% |
| LMArena Search | 1207 | — |
| METR Time Horizons | 75.3% | — |
| Vending-Bench 2 | 3,591 | — |
Reasoning Qwen3.8 Max leads
GPT-5.2: 50.2 (#35), Qwen3.8 Max: 54.4 (#26)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| NYT Connections (extended) | 83.6% | 88.3% |
| Chess Puzzles | 49% | 40% |
| LMArena Hard Prompts | 1445 | 1496 |
| Mystery Game Puzzles | 23% | 38% |
| DTBench | 90.9% | 92% |
| LMCA | 43.9% | 46.2% |
| Epoch Capabilities Index | 153.45 | 156.41 |
| ARC-AGI-2 | 52.9% | — |
| SimpleBench | 45.8% | — |
| Kagi LLM Benchmark | 73.3% | — |
| ARC-AGI-1 | 86.2% | — |
| CritPt | — | 20% |
| EnigmaEval | 10.4% | — |
| EBR-Bench | 23% | — |
| ForecastBench | 60.1 | — |
Math Qwen3.8 Max leads
GPT-5.2: 60.0 (#38), Qwen3.8 Max: 73.2 (#20)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| FrontierMath (Tiers 1-3) | 67.4% | 74.7% |
| FrontierMath Tier 4 | 31.7% | 46.3% |
| OTIS Mock AIME 2024-2025 | 96.1% | 100% |
| ProofBench | 15% | 58% |
| LMArena Math | 1440 | 1499 |
| MathArena Final-Answer Competitions | 72% | — |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Tier 4 (v1) | 18.8% | — |
Knowledge Qwen3.8 Max leads
GPT-5.2: 59.3 (#32), Qwen3.8 Max: 61.7 (#27)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| GPQA Diamond | 91.4% | 92.7% |
| SimpleQA Verified | 37.1% | 47.3% |
| LMArena Expert | 1445 | 1507 |
| Humanity's Last Exam | 27.8% | — |
| Vectara Hallucination Rate | 8.4% | — |
Multimodal GPT-5.2 leads
GPT-5.2: 51.3 (#7), Qwen3.8 Max: 37.2 (#75)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | 1268 | 1314 |
| Furniture Assembly | 38.3% | 20% |
| VPCT | 84% | — |
| LMArena Document | 1405 | — |
Multilingual Qwen3.8 Max leads
GPT-5.2: 53.4 (#67), Qwen3.8 Max: 56.7 (#18)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1425 | 1472 |
| LMArena Chinese | 1460 | 1538 |
| LMArena French | 1455 | 1503 |
| LMArena German | 1448 | 1483 |
| LMArena Japanese | 1420 | 1467 |
| LMArena Korean | 1392 | 1461 |
| LMArena Russian | 1440 | 1481 |
| LMArena Spanish | 1433 | 1492 |
Instruction Following Qwen3.8 Max leads
GPT-5.2: 74.7 (#89), Qwen3.8 Max: 77.6 (#17)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1417 | 1479 |
Long Context Qwen3.8 Max leads
GPT-5.2: 44.0 (#78), Qwen3.8 Max: 45.6 (#31)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1428 | 1489 |
| CL-bench | 18.2% | — |
Writing & Preference Too close to call
GPT-5.2: 66.8 (#32), Qwen3.8 Max: 67.1 (#30)
| Benchmark | GPT-5.2 | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1439 | 1483 |
| LMArena Creative Writing | 1401 | 1479 |
| LMArena Multi-Turn | 1458 | 1489 |
| EQ-Bench Creative Writing | 1703 | — |
Frequently asked questions
Is GPT-5.2 better than Qwen3.8 Max?
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 54.1 on the Noometry Index.
Which is cheaper, GPT-5.2 or Qwen3.8 Max?
Qwen3.8 Max is cheaper. It lists at $2 per million input tokens and $6 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GPT-5.2 or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 51.6 in the Noometry coding category.
Which has the bigger context window?
Qwen3.8 Max does, with 1M tokens against 400K.
How many benchmarks do GPT-5.2 and Qwen3.8 Max share?
33 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and Qwen3.8 Max has 39.