Model comparison
GPT-6 Astra vs Qwen3 235B-A22B
GPT-6 Astra is the stronger model overall, scoring 70.8 to 43.5 on the Noometry Index. Qwen3 235B-A22B costs 16× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 33 shared benchmarks.
Alibaba (Qwen)Summary
- They share 33 benchmarks with published results for both. GPT-6 Astra scores higher in 8 categories and Qwen3 235B-A22B in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 15.7.
- The biggest single-benchmark swing is ARC-AGI-2: 95% for GPT-6 Astra and 1.3% for Qwen3 235B-A22B.
- Qwen3 235B-A22B is cheaper at $0.70 / $2.80 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 131K.
- Qwen3 235B-A22B has downloadable open weights; the other is API-only.
Side by side
| GPT-6 Astra | Qwen3 235B-A22B | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 70.8 | 43.5 |
| Released | 2026-09-03 | 2025-04 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 131K |
| Max output | 128K | 16K |
| Input $ / M tokens | $10 | $0.70 |
| Output $ / M tokens | $50 | $2.80 |
| Results tracked | 56 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-6 Astra: 73.7 (#2), Qwen3 235B-A22B: 44.3 (#75)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| SciCode | 56.5% | 42.4% |
| WeirdML | 93.6% | 41% |
| LMArena Coding | 1487 | 1445 |
| DeepSWE | 74.1% | — |
| FrontierCode | 53.3% | — |
| Aider Polyglot | — | 59.6% |
| LMArena WebDev | 1786 | — |
| FrontierSWE | 65.5% | — |
| GSO | 79.4% | — |
| MirrorCode | 46.7% | — |
| ALE-Bench | 2,951 | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.9 (#3), Qwen3 235B-A22B: 33.9 (#51)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| Vending-Bench 2 | 15,515 | -11.34 |
| APEX-Agents | 64.7% | — |
| Berkeley Function Calling Leaderboard | — | 52.1% |
| Remote Labor Index | 20.8% | — |
| BALROG | 68.3% | — |
| GDP.pdf | 34.2% | — |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), Qwen3 235B-A22B: 15.7 (#311)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| ARC-AGI-2 | 95% | 1.3% |
| ARC-AGI-1 | 98.5% | 11% |
| CritPt | 31.7% | 0% |
| Chess Puzzles | 72% | 12% |
| LMArena Hard Prompts | 1462 | 1433 |
| Mystery Game Puzzles | 84% | 9% |
| DTBench | 97.3% | 80.3% |
| LMCA | 64.4% | 29.3% |
| Epoch Capabilities Index | 166.45 | 143.85 |
| SimpleBench | — | 31% |
| Kagi LLM Benchmark | — | 69.4% |
| NYT Connections (extended) | 98.1% | — |
| EBR-Bench | 76.2% | — |
| Bench to the Future 3 | 0.14 | — |
| ForecastBench | — | 59.7 |
Math GPT-6 Astra leads
GPT-6 Astra: 93.5 (#2), Qwen3 235B-A22B: 50.4 (#57)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 100% | 86.7% |
| LMArena Math | 1465 | 1432 |
| FrontierMath (Tiers 1-3) | 93.7% | — |
| FrontierMath Tier 4 | 97.6% | — |
| ProofBench | 99% | — |
| Omni-MATH | — | 71.8% |
| MATH Level 5 | — | 68.9% |
| FrontierMath (Feb 2025 set) | — | 8.5% |
| FrontierMath Erdős | 2.9% | — |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge GPT-6 Astra leads
GPT-6 Astra: 75.3 (#1), Qwen3 235B-A22B: 49.6 (#73)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| GPQA Diamond | 95.8% | 80.1% |
| SimpleQA Verified | 75.6% | 40.4% |
| Vectara Hallucination Rate | 8.7% | 9.3% |
| LMArena Expert | 1483 | 1463 |
| Humanity's Last Exam | 54.8% | — |
| MMLU-Pro | — | 84.4% |
| Confabulations | — | 15.6% |
| GPQA (HELM) | — | 72.7% |
Multimodal Not comparable
GPT-6 Astra: 55.0 (#3), Qwen3 235B-A22B: —
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| LMArena Vision | 1281 | — |
| Blueprint-Bench 2 | 49.7% | — |
| Furniture Assembly | 80% | — |
| LMArena Document | 1468 | — |
Multilingual GPT-6 Astra leads
GPT-6 Astra: 53.7 (#61), Qwen3 235B-A22B: 52.3 (#89)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| LMArena Non-English | 1430 | 1409 |
| LMArena Chinese | 1484 | 1481 |
| LMArena French | 1456 | 1445 |
| LMArena German | 1440 | 1433 |
| LMArena Japanese | 1379 | 1399 |
| LMArena Korean | 1426 | 1391 |
| LMArena Russian | 1436 | 1411 |
| LMArena Spanish | 1407 | 1430 |
Instruction Following GPT-6 Astra leads
GPT-6 Astra: 76.3 (#44), Qwen3 235B-A22B: 72.6 (#136)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| LMArena Instruction Following | 1450 | 1408 |
| IFEval | — | 83.5% |
Long Context Qwen3 235B-A22B leads
GPT-6 Astra: 44.5 (#62), Qwen3 235B-A22B: 46.1 (#26)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| LMArena Longer Query | 1456 | 1426 |
| Fiction.LiveBench | — | 75% |
Writing & Preference GPT-6 Astra leads
GPT-6 Astra: 75.3 (#7), Qwen3 235B-A22B: 59.6 (#108)
| Benchmark | GPT-6 Astra | Qwen3 235B-A22B |
|---|---|---|
| LMArena Text | 1441 | 1419 |
| LMArena Creative Writing | 1418 | 1384 |
| EQ-Bench Creative Writing | 2173 | 1366 |
| LMArena Multi-Turn | 1448 | 1432 |
| Short-Story Creative Writing | — | 83% |
| WildBench | — | 86.6% |
Frequently asked questions
Is GPT-6 Astra better than Qwen3 235B-A22B?
GPT-6 Astra is the stronger model overall, scoring 70.8 to 43.5 on the Noometry Index. Qwen3 235B-A22B costs 16× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Which is cheaper, GPT-6 Astra 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-6 Astra lists at $10 and $50.
Is GPT-6 Astra or Qwen3 235B-A22B better for coding?
GPT-6 Astra scores higher on coding benchmarks: 73.7 versus 44.3 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 131K.
How many benchmarks do GPT-6 Astra and Qwen3 235B-A22B share?
33 benchmarks have published results for both models. GPT-6 Astra has 56 scored results on Noometry and Qwen3 235B-A22B has 49.