Model board / Qwen3.8-Max

Verified 2026-09-03 against the pages listed below. Prices are USD per 1M tokens, list rate — no committed-use, enterprise or volume discount.

Model Studio Singapore/international rate. The page also shows a region-varying range ($1.65–$2.00 in / $0.206–$0.25 cached / $4.951–$6.00 out); this row uses the international figures for consistency with the rest of the board. No promo/expiry stated on the pricing page, unlike Qwen3.7-Max's

the API states max input 991,808 + max output 131,072; the released weights' native training context is 262,144, extensible to about 1,010,000 per the Hugging Face card — the API figure and the checkpoint's native figure are not the same number

explicit cache: write 125% of input, read 10%, minimum 1,024 tokens. Implicit cache reads at 20% and is not guaranteed to hit; the two are mutually exclusive

Licence: custom "qwen3.8-max" licence named on the Hugging Face model card — not Apache-2.0; the card names the licence but does not print its terms, re-checked 2026-09-03

Watch out

Ends a 3-pass deferral on this board, but the API's vision and video input are not in the Hugging Face release — the open checkpoint is text-only despite the hosted API being multimodal. Read the named custom licence before assuming Apache-2.0-style terms

What we would use it for

Alibaba's new flagship — a 2.4T-parameter MoE (95B active), now with open text-only weights

What this has cost, week by week

Output price at each refresh since 2026-08-20. A vendor change is the provider changing its price. Our correction means this board changed what it prints and the provider's charge did not move — the two are never merged, because publishing the second as the first would be a price rise that never happened.

Nothing has moved since we started tracking it.

Where these numbers came from

• https://www.alibabacloud.com/help/en/model-studio/qwen3-8-max
• https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B

No number on this page was written that was not read from one of those pages on 2026-09-03. If one is wrong, send us the source URL and it is corrected in the next weekly refresh, with a changelog line saying what was wrong — contact@naderu.com.

Jobs this is one of our picks for

• Document understanding — OCR, scans, charts and forms (document-understanding)

Discussed in Naderu Weekly

• The weights aren't the whole model — 2026-08-20

The rest of the board

This row sits alongside 53 others on the model board, refreshed every week. The same data as JSON, and the terms for using it, are at /board-data.