Alibaba introduces Qwen 3.8 Max with API access
Alibaba launched Qwen 3.8 Max with global API access and plans for model weights. The launch announcement did not verify the pricing quoted in the earlier article.
Source update added October 1, 2026. Alibaba Cloud’s August 3 announcement confirms Qwen 3.8 Max, with 2.4 trillion total parameters, 95 billion active parameters, and a context window up to one million tokens.
A preview introduces a model. A launch should also explain where users can access it and which capabilities have supporting evidence. Those are separate questions.
From preview to launch
The announcement confirmed API access through Alibaba Cloud Model Studio for global developers and access through QwenWork. That provides a firmer availability basis than the earlier preview.
The earlier Qwen preview and the August launch remain separate events.
How to read the reported results
Alibaba reported results for coding, visual tasks, and extended agent work. These are company reported findings. This article does not independently reproduce those tests or establish a general performance ranking.
For an evaluation that matters to a deployment, keep the task, inference settings, sampling budget, and scoring method consistent. A model can do well on one test without being the best option for every workflow.
Pricing and access
Correction added October 1, 2026. The $2 input and $6 output figures previously repeated from a tracker were not verified in the primary announcement reviewed. They should not be treated as a confirmed current quote.
Plans for model weights
Alibaba scheduled model weights for the following week in its launch announcement. A planned release should not be described as weights already available on August 3.
The distinction matters for reproducibility and deployment planning. API availability does not by itself establish that a team can download and run the same model locally.
Before budgeting a deployment, check the current endpoint, service terms, usage limits, and price for the specific region and model. None of those details should be inferred from another vendor’s launch.
The announcement supports the model identity and API launch. Pricing and any later weights release need their own dated evidence.
[…] earlier coverage of Qwen 3.8 Max API access similarly separates a hosted service release from plans for downloadable […]