Qwen 3.8-27B watches hours of video on your own computer.
Most models this size can't even see. Everyone treats it like a plain chatbot. That's missing the best part. Here's what the small open model does: → Hand it photos, charts, scanned docs, or full recordings → Drop in
Qwen 3.8-27B is now on Ollama.
One line in your terminal and it's yours forever. Type "ollama run qwen3.8" and a frontier-level agent downloads to your machine. Here's what you just installed: → It watches video. Hand it an hour-long meeting and ask what you promised → A
Everyone is comparing Gemini 3.7 Flash vs Qwen 3.8 27B on benchmarks.
I think that misses the useful question. What happens when you use BOTH models inside the same business workflow? That setup is far more interesting.
Qwen 3.8 Max ran an autonomous coding project for 16 days straight.
No human prompts. 265 commits. 127 pull requests. It took feedback, fixed itself, and kept going. Here's why it's the agent king right now: → It beat Fable 5 on OS World: 86.1 vs 85.0 → That test measures
You have reached the end of the archive
All of qwen38