GPT-6 Astra vs Qwen3.5 397B A17B
GPT-6 Astra 
Qwen3.5 397B A17B Verdict
- GPT-6 Astra (max) leads on quality: 70.9 vs 58.7.
- GPT-6 Astra (max) is stronger in coding, composite, knowledge, maths, reasoning.
- Qwen3.5 397B A17B is stronger in agents & tools, human preference, instruction following, long context, multimodal.
- Qwen3.5 397B A17B is 52× cheaper ($0.387 vs $20.00 per 1M blended).
- Qwen3.5 397B A17B streams 2.9× faster (78 vs 27 tokens per second).
| Metric | GPT-6 Astra (max) | Qwen3.5 397B A17B |
|---|---|---|
| BenchLeader Index | 70.9 | 58.7 |
| Agents & tools score | 68.2 | 69.4 |
| Coding score | 78.6 | 51.7 |
| Composite score | 72.9 | 53.1 |
| Knowledge score | 80.0 | 49.5 |
| Maths score | 77.2 | 50.1 |
| Reasoning score | 73.6 | 63.0 |
| Human preference score | – | 63.6 |
| Instruction following score | – | 76.1 |
| Long context score | – | 65.2 |
| Multimodal score | – | 61.6 |
| Blended price $/M | $20.00 | $0.387 |
| Output speed | 27 tok/s | 78 tok/s |
| Time to first answer | 337.3 s | 42.7 s |
| Context window | 1.1M | 262k |
| GPQA Diamond | 95.8% | 85.9% |
| FrontierMath Tiers 1–3 | 93.7% | 29.5% |
| FrontierMath Tier 4 | 97.6% | – |
| OTIS Mock AIME | 100.0% | 88.9% |
| SimpleQA Verified | 75.6% | – |
| Terminal-Bench | 58.2% | – |
| SciCode | 56.5% | – |
| FrontierCode | 53.3% | – |
| Epoch Capabilities Index | – | 147.0 |
| LMArena Text | – | 1441 |
| LMArena Hard Prompts | – | 1463 |
| LMArena Coding | – | 1491 |
| LMArena WebDev | 1796 | 1399 |
| LMArena Vision | – | 1265 |
| LMArena Agent | 12.5 | – |
| LiveBench | 82.2% | – |
| LiveBench Reasoning | 92.7% | – |
| LiveBench Coding | 80.4% | – |
| LiveBench Agentic Coding | 57.3% | – |
| LiveBench Mathematics | 96.8% | – |
| LiveBench Data Analysis | 83.0% | – |
| LiveBench Language | 89.4% | – |
| AA Intelligence Index | – | 19.1 |
| IFBench | – | 78.8% |
| AA-LCR | – | 77.3% |
| MMMU-Pro | – | 77.3% |
| AA-Omniscience | – | -30.8 |
| Terminal-Bench Hard | – | 40.9% |
| GPQA Diamond (AA) | – | 89.3% |
| Humanity's Last Exam (AA) | – | 29.0% |
| SciCode (AA) | – | 44.8% |
| τ²-Bench Telecom (AA) | – | 95.6% |
| Terminal-Bench 2.1 (Vals) | 87.3% | – |
| Vals Index | 66.6 | – |
| AIME 2026 | – | 94.2% |
| HMMT February 2026 | – | 87.9% |
| ARC-AGI-1 | 97.5% | – |
| ARC-AGI-2 | 95.0% | – |
| ARC-AGI-3 | 98.6% | – |
Data as of 2026-09-10. Best configuration of each model; every score links to its source on the model pages.
GPT-6 Astra vs Qwen3.5 397B A17B: questions
- Is GPT-6 Astra better than Qwen3.5 397B A17B?
- GPT-6 Astra (max) leads on quality: 70.9 vs 58.7. The BenchLeader Index combines every independent quality benchmark; GPT-6 Astra (max) is ahead overall as of 2026-09-10, but check the category scores for your use.
- Is GPT-6 Astra better than Qwen3.5 397B A17B for coding?
- GPT-6 Astra scores higher in coding (79 vs 52 on the category index, where 50 is average).
- Is GPT-6 Astra better than Qwen3.5 397B A17B for agentic tasks?
- Qwen3.5 397B A17B scores higher in agentic tasks (69 vs 68 on the category index, where 50 is average).
- Which is cheaper, GPT-6 Astra or Qwen3.5 397B A17B?
- Qwen3.5 397B A17B is cheaper: $0.387 against $20.00 per million tokens, blended at three input tokens per output token.
- Which is faster, GPT-6 Astra or Qwen3.5 397B A17B?
- Qwen3.5 397B A17B streams faster: 78 against 27 output tokens per second.
- Which has the larger context window?
- GPT-6 Astra accepts more context: 1.1M against 262k tokens.