Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Allow fast_inference for Qwen3.5 (#10891)

Daniel Hanf8cd9149eeec
diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py
index ba5e2f872db..17429f7283e 100644
--- a/unsloth/models/vision.py
+++ b/unsloth/models/vision.py
@@ -549,6 +549,9 @@ VLLM_SUPPORTED_VLM = [
"mistral3",
"qwen3_vl",
"qwen3_vl_moe",
+ # Qwen3.5 ships as Qwen3_5ForConditionalGeneration with a vision_config, so it
+ # reaches this gate even for the text-only checkpoints.
+ "qwen3_5",
]
VLLM_NON_LORA_VLM = [
"mllama",