[Bug] No Model Detected Error after Hitting Send

[TheAlex25]

Greetings. After the new update, Unsloth broke itself. When pressing enter for a new prompt in a new session if no model is loaded into memory, instead of loading my default model, Gemma-4-26B-A4B, Unsloth decides to say No Model Loaded, Loading Qwen 3.5 4B. This did not happen with the previous version of Unsloth, as the previous version would detect my downloaded Gemma GGUF, load it into memory, and wait to run the prompt until the model is in RAM/VRAM.

First off, Unsloth should not autofetch a model off of HuggingFace without my permission, and secondly, Gemma 4 is on my SSD. Unsloth should have recognized that model has existed.

Error code: "Downloading a small model… No downloaded models found. Fetching Qwen3.5-4B-MTP (UD-Q4_K_XL)."

Unsloth Version
v0.1.501-beta
Package Version
2026.7.5
llama.cpp Version
b10079-mix-fb3d4ca

Specs

System Details Report

---

Report details

• Date generated: 2026-07-23 12:12:11

Hardware Information:

• Hardware Model: Lenovo Yoga Pro 9 16IMH9
• Memory: 32.0 GiB
• Processor: Intel® Core™ Ultra 9 185H × 22
• Graphics: Intel® Arc™ Graphics (MTL)
• Graphics 1: NVIDIA GeForce RTX™ 4050 Laptop GPU
• Disk Capacity: 1.0 TB

Software Information:

• Firmware Version: NKCN28WW
• OS Name: Ubuntu 26.04 LTS
• OS Build: (null)
• OS Type: 64-bit
• GNOME Version: 50
• Windowing System: Wayland
• Kernel Version: Linux 7.0.0-27-generic

Screenshot