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 reads more at once: a context window of 1M tokens against 262K.
- Qwen3.8-2.4T-A95B can be downloaded and run on your own machine; Qwen3.8-Max can't.
- No independent board we follow ranks both models yet.
Side by side
The specs.
| Qwen3.8-2.4T-A95B | Qwen3.8-Max | |
|---|---|---|
| Made by | Alibaba (Qwen) · China | Alibaba (Qwen) · China |
| Released | Aug 8, 2026 | Sep 2, 2026 |
| Price per 1M tokens | Free to download; hosts set prices | $2 in / $6 out |
| Context window | 262K | 1M |
| Takes in | text | text, image, video |
| Download and run it yourself | Yes (qwen3.8-max (license tag shown on Hugging Face)) | No |
| Size | 2.4T parameters, 95B active | 2.4T (MoE) parameters, 95B active |
| Maker says | “For the first time, Qwen3.8 brings a Qwen-Max-class model to open release.” | “A 2.4-trillion-parameter MoE flagship with a major leap in coding and office productivity” |
| Sources | Prices, Oct 11, 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 | Qwen3.8-2.4T-A95B | Qwen3.8-Max |
|---|---|---|
| Epoch Capabilities IndexEpoch AI, Oct 20, 2026 | Not listed | #20156.88 |
| LMArena creative writingLMArena, Oct 15, 2026 | Not listed | #19 |
| LMArena WebDevLMArena, Oct 19, 2026 | Not listed | #11 |
| LMArena mathLMArena, Oct 15, 2026 | Not listed | #18 |
| LMArena visionLMArena, Oct 19, 2026 | Not listed | #2 |
| Terminal-Bench 4.0Vals AI, Oct 19, 2026 | Not listed | #16 |
| 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 |
More comparisons