Qwen3.8-Max moves to conditional: an official $2/$6 per 1M token rate, benchmarks still missing
- What happened
- Alibaba Cloud Model Studio's official pricing page lists Qwen3.8-Max at $2 input and $6 output per 1M tokens (Singapore).
- Why it matters
- A verified per-token price makes it worth a paid evaluation, so we move it to conditional; benchmarks, a model card and confirmed open weights are still absent.
- What to do
- Evaluate it on your own tasks at the listed rate; do not commit production workloads without independent benchmarks.
Verdict: Qwen3.8-Max moves from pending to conditional. Alibaba Cloud Model Studio's official pricing page lists a flat per-token rate of $2 input and $6 output per 1M tokens (Singapore), so the model can now be costed against an official rate and is worth a paid evaluation. It is not yet a production pick: there is still no benchmark table, no model card and no confirmed open-weight release.
What happened
Alibaba Cloud Model Studio's pricing page lists Qwen3.8-Max at:
| Region | Input (per 1M tokens) | Output (per 1M tokens) | Notes |
|---|---|---|---|
| Singapore (International) | $2 | $6 | One tier up to 1M tokens, context caching discount, 50% batch discount, 1M free tokens (90 days) |
| China (Beijing) | $1.65 | $4.951 | One tier up to 1M tokens, context caching discount, 50% batch discount, no free quota |
Batch and context caching discounts cannot be combined.
The International listing supports both thinking and non-thinking modes. Model Studio's model list places Qwen3.8-Max under text generation ("complex reasoning and coding") and under image and video understanding.
Why it matters
- Price is verified on an official page. Model Studio's pricing page lists a flat $2/$6 per 1M tokens (Singapore), so you can budget a real test against an official rate.
- Quality is still unverified. We found no published benchmark table or model card, and neither Model Studio page we read carries benchmark numbers or comparisons to other models.
- Weights are unconfirmed. We could not read Alibaba's qwen.ai blog pages, so we make no claim either way about an open-weight release or license. If you need a Qwen 3.8 model you can self-host today, that is Qwen3.8-27B, not Max.
What changes for you
- Already on Alibaba Cloud or the Qwen ecosystem: run Qwen3.8-Max on your own coding and reasoning tasks at the listed rate. The 1M free tokens on the International listing (valid 90 days) cover a first pass.
- Planning a production migration: hold until independent benchmarks or a model card appear. Conditional means worth evaluating, not worth committing to.
- Comparing costs: use the region you will actually deploy in. Beijing rates are lower than Singapore rates.
Qwen3.8-Max
Alibaba Cloud Model Studio's official pricing page lists $2/$6 per 1M tokens (Singapore); benchmarks, a model card and open weights remain unconfirmed.
Affected tools & models
Never need to catch up again
The weekly delta: only verdict changes and act-now items. No digest filler.