In short
How they differ.
Worked out from the figures below. Prices and specs are the makers' own, read Oct 20, 2026.
- Qwen3.8-Max is cheaper per output token: $6 against $50 per million tokens, so GPT-6 Astra costs 8.3 times as much.
- GPT-6 Astra reads more at once: a context window of 1M tokens against 1M.
- On the 3 independent boards that rank both, GPT-6 Astra places higher on 3 and Qwen3.8-Max on 0.
Side by side
The specs.
| GPT-6 Astra | Qwen3.8-Max | |
|---|---|---|
| Made by | OpenAI · US | Alibaba (Qwen) · China |
| Released | Sep 3, 2026 | Sep 2, 2026 |
| Price per 1M tokens | $10 in / $50 out | $2 in / $6 out |
| Context window | 1M | 1M |
| Takes in | text, image | text, image, video |
| Download and run it yourself | No | No |
| Size | Not published | 2.4T (MoE) parameters, 95B active |
| Maker says | “Our most capable model for the most demanding work.” | “A 2.4-trillion-parameter MoE flagship with a major leap in coding and office productivity” |
| Sources | Prices, Oct 20, 2026 · Description | Prices, Oct 11, 2026 · Description |
Independent tests
Where they place.
Position on each board we follow (1 is best). Epoch AI's and Vectara's scores are shown; the other boards are linked for their own figures.
| Board | GPT-6 Astra | Qwen3.8-Max |
|---|---|---|
| Epoch Capabilities IndexEpoch AI, Oct 20, 2026 | #2166.41 | #20156.88 |
| LMArena creative writingLMArena, Oct 15, 2026 | Not listed | #19 |
| LMArena coding chatsLMArena, Oct 15, 2026 | #9 | Not listed |
| LMArena WebDevLMArena, Oct 19, 2026 | #2 | #11 |
| LMArena mathLMArena, Oct 15, 2026 | Not listed | #18 |
| LMArena visionLMArena, Oct 19, 2026 | Not listed | #2 |
| LMArena documentsLMArena, Sep 13, 2026 | #16 | Not listed |
| LMArena Agent ArenaLMArena, Oct 19, 2026 | #2 | Not listed |
| Humanity's Last ExamScale AI | #1 | Not listed |
| SWE-Bench ProScale AI, Sep 22, 2026 | #5 | Not listed |
| Remote Labor IndexScale AI | #2 | Not listed |
| Vals IndexVals AI, Oct 19, 2026 | #5 | Not listed |
| Terminal-Bench 4.0Vals AI, Oct 19, 2026 | #3 | #16 |
| Vibe Code BenchVals AI, Oct 18, 2026 | #7 | Not listed |
| ProofBenchVals AI, Oct 12, 2026 | #6 | Not listed |
| GPQA Diamond (run by Vals AI, archived)Vals AI, Sep 1, 2026 | Not listed | #6 |
| MMMU Pro (run by Vals AI, archived)Vals AI, Sep 1, 2026 | Not listed | #12 |
| CUA-bench (games)Vals AI, Oct 13, 2026 | #1 | Not listed |
| Vectara hallucination testVectara, Oct 16, 2026 | #458.7% | Not listed |
More comparisons