🚀 FastFlowLM v0.9.37 - New Qwen3.5 Models + Better Defaults

FastFlowLM v0.9.37 adds two new Qwen3.5 vision-language models and improves the default model path experience for new Windows users.

---

🖼️ 1. New Qwen3.5 Models

FastFlowLM now supports two additional vision-language models from the Qwen3.5 family:

• Tag: qwen3.5:2b
• Tag: qwen3.5:9b

You can control image pre-resizing in both run and serve modes with --img-pre-resize (or -r).

⚠️ Note: The default pre-resize setting for all Qwen VL models (2.5, 3 and 3.5) is now 720p.

Run in CLI mode:

flm run qwen3.5:2b -r 1

Run in server mode:

flm serve qwen3.5:9b -r 1

For more details, see the model card (2B & 9B) and benchmark results.

---

🪟 2. Better Default Model Path on Windows

For new Windows users, FLM_MODEL_PATH now defaults to C:\Users\<username>\.flm.

Existing users are unaffected: the previous default path, C:\Users\<username>\Documents\flm, is still supported, so current model installations continue to work without changes.

---

🌟 Summary

FastFlowLM v0.9.37 expands the Qwen3.5 lineup with qwen3.5:2b and qwen3.5:9b, while improving the out-of-box experience for new Windows users with a cleaner default model path.