Alibaba Cloud Model Studio’s docs describe a limited-time night Credits softening for Token Plan Personal Edition when listed models run in the overnight window. It adjusts how many Credits those named models consume. It does not rewrite other models’ rates, turn Personal into Team, or cut the monthly subscription sticker price.

The window is every night from 22:00 to 08:00 (UTC+8). On the international docs page, Personal Edition is currently available in the Singapore region.

What’s discounted (official)

  • qwen3.8-max and qwen3.8-flash: 60% off Credits consumption in the night window (Chinese docs say 「4 折」—same idea: 40% of the usual Credits burn).
  • deepseek-v4-pro-0813, deepseek-v4-flash-0731, deepseek-v4.1-flash, and deepseek-v4.1-flash-1017: 50% off Credits consumption in the same window (「5 折」).
  • decision-model-preview: the China-site docs list it as limited-time free (0 Credits); the international English overview does not list that line yet.
  • qwen3.8-max-preview is retired: the old model ID still works, requests route to qwen3.8-max, and Credits／usage count at the max rate. The config model ID should be updated to qwen3.8-max when possible.

Alibaba Cloud Model Studio says it may change or adjust the promo—including content and validity—based on operations. Use the latest page or Alibaba Cloud notices; the docs do not nail a fixed end date.

The same overview says Personal Edition works with mainstream coding／agent tools such as Claude Code, Cursor, and Qwen Code. That is compatibility wording, not an unlimited-use promise.

Real overnight Credits burn on these IDs has not been measured on this site—for reference only. Measured results can go through the site-owner contact link in the footer.

Related reading: Volcengine Ark Coding Plan Softens Deduction Rates for Kimi and DeepSeek Models.