Model comparison
GPT-6 Astra vs Qwen2.5-VL 72B Instruct
GPT-6 Astra is the stronger model overall, scoring 70.8 to 29.9 on the Noometry Index. Qwen2.5-VL 72B Instruct costs 4.8× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 1 shared benchmarks.
Alibaba (Qwen)Summary
- They share 1 benchmark with published results for both. GPT-6 Astra scores higher in 3 categories and Qwen2.5-VL 72B Instruct in 0 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 20.7.
- Qwen2.5-VL 72B Instruct is cheaper at $2.80 / $8.40 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 131K.
- Qwen2.5-VL 72B Instruct has downloadable open weights; the other is API-only.
Side by side
| GPT-6 Astra | Qwen2.5-VL 72B Instruct | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 70.8 | 29.9 |
| Released | 2026-09-03 | 2024-09 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 131K |
| Max output | 128K | 8K |
| Input $ / M tokens | $10 | $2.80 |
| Output $ / M tokens | $50 | $8.40 |
| Results tracked | 56 | 6 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Not comparable
GPT-6 Astra: 73.7 (#2), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| DeepSWE | 74.1% | — |
| FrontierCode | 53.3% | — |
| LMArena WebDev | 1786 | — |
| FrontierSWE | 65.5% | — |
| SciCode | 56.5% | — |
| GSO | 79.4% | — |
| WeirdML | 93.6% | — |
| LMArena Coding | 1487 | — |
| MirrorCode | 46.7% | — |
| ALE-Bench | 2,951 | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.9 (#3), Qwen2.5-VL 72B Instruct: 18.6 (#144)
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| APEX-Agents | 64.7% | — |
| Remote Labor Index | 20.8% | — |
| OSWorld | — | 5% |
| BALROG | 68.3% | — |
| GDP.pdf | 34.2% | — |
| Vending-Bench 2 | 15,515 | — |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), Qwen2.5-VL 72B Instruct: 20.7 (#233)
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| ARC-AGI-2 | 95% | — |
| Kagi LLM Benchmark | — | 36% |
| NYT Connections (extended) | 98.1% | — |
| ARC-AGI-1 | 98.5% | — |
| CritPt | 31.7% | — |
| Chess Puzzles | 72% | — |
| EBR-Bench | 76.2% | — |
| LMArena Hard Prompts | 1462 | — |
| Mystery Game Puzzles | 84% | — |
| DTBench | 97.3% | — |
| LMCA | 64.4% | — |
| Bench to the Future 3 | 0.14 | — |
| Epoch Capabilities Index | 166.45 | — |
Math Not comparable
GPT-6 Astra: 93.5 (#2), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| FrontierMath (Tiers 1-3) | 93.7% | — |
| FrontierMath Tier 4 | 97.6% | — |
| OTIS Mock AIME 2024-2025 | 100% | — |
| ProofBench | 99% | — |
| LMArena Math | 1465 | — |
| FrontierMath Erdős | 2.9% | — |
Knowledge Not comparable
GPT-6 Astra: 75.3 (#1), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| GPQA Diamond | 95.8% | — |
| Humanity's Last Exam | 54.8% | — |
| SimpleQA Verified | 75.6% | — |
| Vectara Hallucination Rate | 8.7% | — |
| LMArena Expert | 1483 | — |
Multimodal GPT-6 Astra leads
GPT-6 Astra: 55.0 (#3), Qwen2.5-VL 72B Instruct: 33.5 (#97)
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Vision | 1281 | 1107 |
| Video-MME | — | 73.5% |
| GeoBench | — | 62% |
| Blueprint-Bench 2 | 49.7% | — |
| Furniture Assembly | 80% | — |
| LMArena Document | 1468 | — |
| SpatialViz-Bench | — | 33.3% |
Multilingual Not comparable
GPT-6 Astra: 53.7 (#61), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Non-English | 1430 | — |
| LMArena Chinese | 1484 | — |
| LMArena French | 1456 | — |
| LMArena German | 1440 | — |
| LMArena Japanese | 1379 | — |
| LMArena Korean | 1426 | — |
| LMArena Russian | 1436 | — |
| LMArena Spanish | 1407 | — |
Instruction Following Not comparable
GPT-6 Astra: 76.3 (#44), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1450 | — |
Long Context Not comparable
GPT-6 Astra: 44.5 (#62), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1456 | — |
Writing & Preference Not comparable
GPT-6 Astra: 75.3 (#7), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-6 Astra | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Text | 1441 | — |
| LMArena Creative Writing | 1418 | — |
| EQ-Bench Creative Writing | 2173 | — |
| LMArena Multi-Turn | 1448 | — |
Frequently asked questions
Is GPT-6 Astra better than Qwen2.5-VL 72B Instruct?
GPT-6 Astra is the stronger model overall, scoring 70.8 to 29.9 on the Noometry Index. Qwen2.5-VL 72B Instruct costs 4.8× 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 Qwen2.5-VL 72B Instruct?
Qwen2.5-VL 72B Instruct is cheaper. It lists at $2.80 per million input tokens and $8.40 per million output tokens; GPT-6 Astra lists at $10 and $50.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 131K.
How many benchmarks do GPT-6 Astra and Qwen2.5-VL 72B Instruct share?
1 benchmark has published results for both models. GPT-6 Astra has 56 scored results on Noometry and Qwen2.5-VL 72B Instruct has 6.