[Models] Qwen3 Embedding shown as LLM instead of an embedding model

[YorkieDev]

Which version of LM Studio?
0.3.16(Build 8)

What is the bug?
The New Qwen 3 Embedding Models are labelled as an LLM in the model explorer
https://huggingface.co/collections/Qwen/qwen3-embedding-6841b2055b99c44d9a4c371f

Screenshots

To Reproduce
Steps to reproduce the behavior:

1. Download the newest Qwen3 Embedding Model - Qwen/Qwen3-Embedding-4B-GGUF
2. Navigate to your My Models Tab
3. Observe that it's labelled as an LLM instead of an Embedding Model

[xenv]

In the latest version, I found that we can run it by setting the Domain Type, but after running it, it still returns abnormal results. Different inputs always output the same content, which may be related to the version of llama.cpp. I hope the official can try to support it

[kaleyroy]

i met the same issue +1

[8times4]

Pretty sure once ggml-org/llama.cpp#14029 is fixed, it’ll work properly in lmstudio as well.

[ashunaveed]

same here for Qwen/Qwen3-Embedding-8B-GGUF.

[YorkieDev]

This seems to have been solved in v0.3.17(Build 11) and latest runtime updates.