Head to head
Qwen3.8-Max vs Qwen3-Max
Qwen3.8-Max from Alibaba Qwen against Qwen3-Max from Alibaba Qwen — specification, price and every benchmark both makers have published, in one table.
At a glance
Benchmarks
Qwen3.8-Max leads
Qwen3.8-Max wins 1 of the 1 benchmarks both models report, Qwen3-Max wins 0, consistently. The average gap across shared tests is 7.2 points.
Price
Qwen3-Max is cheaper
On a 3:1 input-to-output mix, Qwen3-Max costs $2.40 per million tokens against $3.00 for Qwen3.8-Max — about 1.2× less. Remember that a reasoning model bills its thinking as output, so cost per answer can diverge much further than cost per token.
What actually differs
- Qwen3.8-Max has open weights under Qwen3.8-Max License, so it can run on your own hardware with no per-token cost and no dependency on an API staying available. The other is API-only.
- Qwen3.8-Max takes 1M tokens of context against 262K — 3.8× more room for long documents or a large codebase.
Scorecard
Which is better at what
Maths, coding, reasoning and the rest — one line each, averaged over the benchmarks both models actually report.
| Category | Qwen3.8-Max | Qwen3-Max | Better at this |
|---|---|---|---|
| ReasoningGPQA Diamond | 92.6% | 85.4% | Qwen3.8-Max+7.2 |
| MathsOnly one model reports this | — | — | Not comparable |
| CodingOnly one model reports this | — | — | Not comparable |
| KnowledgeNeither model reports this | — | — | Not comparable |
| MultimodalNeither model reports this | — | — | Not comparable |
| Instruction followingNeither model reports this | — | — | Not comparable |
| Human preferenceNeither model reports this | — | — | Not comparable |
| Categories wonOut of 1 comparable | 1 | 0 | Qwen3.8-Max |
Each category averages only the benchmarks every model here reports, so no one is credited for a test the other did not run. A category with no shared test is marked Not comparable rather than guessed at.
Side by side
Qwen3.8-Max and Qwen3-Max, row by row
| Attribute | Qwen3.8-MaxAlibaba Qwen | Qwen3-MaxAlibaba Qwen |
|---|---|---|
| Specification | ||
| MakerWho built it | Alibaba Qwen | Alibaba Qwen |
| Released | 2026-08 | 2025-09 |
| ParametersTotal, and active per token for a mixture of experts | 2.4T total / 95B active | 1T+ |
| Architecture | MoE | MoE |
| Context windowHow much can go in at once | 1,000,000 tokens | 262,144 tokens |
| Max output | 131,072 tokens | 32,768 tokens |
| Input | Not reported | Text |
| ReasoningSpends extra tokens thinking before it answers | Not reported | Yes |
| Tool calling | Yes | Yes |
| Licence | Qwen3.8-Max License | Proprietary |
| Open weightsCan you download and run it yourself | Yes | No |
| Price | ||
| Input priceUSD per million tokens in | $2.00 | $1.20 |
| Output priceUSD per million tokens out | $6.00 | $6.00 |
| Blended 3:1A 3-in-to-1-out million-token mix — a fairer single number than input price alone | $3.00 | $2.40Cheapest |
| Published benchmarks | ||
| GPQA Diamond198 graduate-level physics, chemistry and biology questions written to be Google-proof. PhD holders in the matching field score about 65%. | ||
| AIME 2025The American Invitational Mathematics Examination — 15 problems, integer answers, no partial credit. A standard test of multi-step maths reasoning. | Not reported | |
| SWE-bench Verified500 human-validated GitHub issues from real Python repositories. The model must produce a patch that makes the project's own tests pass. | Not reported | |
| SWE-bench ProA harder, contamination-resistant successor to SWE-bench Verified, drawn from commercial and copyleft repositories that were never public training data. | Not reported | |
| Terminal-Bench 2.1The 2.1 revision of the terminal agent benchmark. Scores on it are not comparable with Terminal-Bench 2.0 or 4.0 — the task set changed. | Not reported | |
| Links | Full page | Full page |
| Row verified | 2026-10 | 2026-08 |
Where these numbers come from
Every score on this page is a published figure, taken from the model's own card, system card, technical report or release post, or from a public leaderboard. CorX Labs did not run these evaluations. Most are self-reported by the lab that built the model, which means they were produced under that lab's own choice of prompt, scaffold and number of attempts — so treat them as a starting point for a shortlist, not as a settled ranking.
A score someone other than the model's maker measured is marked Independent and names its measurer. Those are the stronger numbers on this page — an outside harness has no reason to flatter anyone — and there are not many of them.
Where a figure has not been published, the cell reads Not reported rather than an estimate. Nothing here is inferred, interpolated or guessed. Each model records the month its row was last checked. Full method and caveats.
Questions
Qwen3.8-Max or Qwen3-Max?
Is Qwen3.8-Max better than Qwen3-Max?
Qwen3.8-Max wins 1 of the 1 benchmarks both models report, Qwen3-Max wins 0, consistently. The average gap across shared tests is 7.2 points.
Which is cheaper, Qwen3.8-Max or Qwen3-Max?
On a 3:1 input-to-output mix, Qwen3-Max costs $2.40 per million tokens against $3.00 for Qwen3.8-Max — about 1.2× less. Remember that a reasoning model bills its thinking as output, so cost per answer can diverge much further than cost per token.
What is the difference between Qwen3.8-Max and Qwen3-Max?
Qwen3.8-Max has open weights under Qwen3.8-Max License, so it can run on your own hardware with no per-token cost and no dependency on an API staying available. The other is API-only. Qwen3.8-Max takes 1M tokens of context against 262K — 3.8× more room for long documents or a large codebase.
