Qwen 3.8-27B actually shipped — Apache 2.0, vision, 262K context
The model we covered twice as announced-but-absent is now downloadable. 27B causal LM with integrated vision, 262K native context extensible to 1M, Apache 2.0. The delay was three days. The pattern it broke is worth more than the model.
We ran this story twice as an absence — once yesterday and once the day before that. The correction is that it shipped: Qwen 3.8-27B, released 15 August under Apache 2.0, a 27-billion-parameter causal language model with integrated vision and a 262K native context window the team says extends to 1M.
Say the obvious thing first. Three days between announcement and artefact is a normal software delay, and we were right to flag the gap and right to close it this fast. The reason a three-day slip was worth two articles is that announced-open-weights-that-never-arrive has been a recurring move this year, and the only way to make announcements mean anything is to check them.
On the model itself, the configuration is the interesting part. Apache 2.0 on a 27B with vision built in is a genuinely permissive release at a size that fits on hardware a small team owns rather than rents. The 262K native context is the number to watch — not because long context is new, but because native and extended are different claims and only the first is a property of the model.
It lands in a week that also produced GLM-5.3, which we covered yesterday, and it lands from a lab that has now put out both a 2.4-trillion-parameter sparse model and this. Alibaba is running the widest release range of anyone: enormous and undownloadable at one end, mid-size and Apache-licensed at the other.
LLM Stats — LLM News Today (August 2026) — AI Model Releases → · AI Release Tracker — Latest AI Model Releases — August 2026 → · The Open Weights — Open-source AI, tracked daily →