Qwen3.8 Max
Qwen released 2026-08-03 proprietary
GLM 5.3 scores higher and costs less — but you would give something up.
+1.4 on the capability index and 28% cheaper ($2.15/M against $3.00/M). What you lose:
- no image, video input
Capability scores do not measure context length, output ceiling or which inputs a model accepts, so a higher score does not mean a drop-in replacement.
Our take
editorial — not a measurementAlibaba's flagship at $2/$6 with LMArena 1481, a 1M window and full multimodal input including video. Priced identically to the open-weight Qwen3.8 2.4T A95B but proprietary, and the only one of the pair with a published cache-write rate. The Qwen line's real advantage is breadth: the family spans a 2.4T-parameter MoE down to 9B models, mostly Apache-licensed, so you can prototype on the hosted flagship and then drop to self-hosted weights in the same family without changing prompt style.
Strengths
- LMArena 1481 at $2/$6
- 1M context, 131k max output
- Multimodal including video input
- Family continuity from flagship down to small open-weight models
Weaknesses
- Proprietary, unlike most of the Qwen family
- Output at $6 is above Grok 4.6 for a lower Arena score
- No AA intelligence index captured
Reach for it when
- Multimodal work with a self-hosted migration path
- Teams already using Qwen open weights
Avoid it if
- You want open weights — use Qwen3.8 2.4T A95B instead
- Grok 4.6 already covers your need at similar cost
Sources: openrouter.aiarena.ai
Every price dimension
| Input | $2.00/M |
|---|---|
| Output | $6.00/M |
| Cached input | $0.250/M |
| Cache write | $2.50/M |
Independent scores
| Intelligence | 58.1 |
|---|---|
| Coding | 71.8 |
| Agentic | 58.4 |
Sources are listed separately rather than averaged. Across the 62 models both have scored they correlate at r = 0.806 — close agreement overall, but four models rank very differently between them, and a blended score would hide exactly those.
Best rankings by task
- webapps #1of 56 1340
- uicomponent #3of 140 1362
- 3d #3of 136 1376
- mobileapps #3of 56 1265
- fullstack #4of 59 1339
- gamedev #7of 144 1340
- codecategories #8of 145 1323
- website #13of 151 1304
Capability
| Context window | 1M |
|---|---|
| Max output | 131K |
| Input modes | text, image, video |
| Tool use | yes |
| Reasoning | always on |
| Open weights | no |
Provenance
| Price source | openrouter.ai |
|---|---|
| Fetched | 2026-08-24 |
| Quality data | verified |
| Cross-checked | vendor page |
Qwen3.8 Max is the flagship model in Alibaba's Qwen3.8 series, the general-availability successor to the Qwen3.8 Max Preview. It is a multimodal reasoning model intended for complex reasoning, visual understanding,...