Alibaba Qwen Praised for Releasing Daily Checkpoints of Qwen 3.8 Preview
Alibaba's Qwen team is releasing daily checkpoint updates for its Qwen 3.8 preview models, allowing users to test and provide feedback on the model's daily evolution. The latest preview version features broad performance gains and significant improvements on the web frontend. This rapid, iterative release strategy represents an open approach to LLM development, giving the community direct access to the model's training progress before the official open-weight release. It fosters collaborative debugging and helps developers understand how models evolve day-by-day. The Qwen team plans to release a more capable, official version of Qwen 3.8 and open-weight it for the public. Users are encouraged to test the preview version, specifically the Qwen3.8-Max-Preview, and report what breaks.
## BACKGROUND
In machine learning, a checkpoint is a saved state of a model during training, allowing developers to resume training or evaluate the model at specific intervals. An open-weight AI model refers to a model whose trained parameters (weights) are publicly shared, enabling anyone to run, customize, or fine-tune it.