Skip to content

Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.

qwen3vl: enable flash attention by default (#12862)

Michael Yang75e75d9afea9
diff --git a/fs/ggml/ggml.go b/fs/ggml/ggml.go
index c0ca068ab..909104c64 100644
--- a/fs/ggml/ggml.go
+++ b/fs/ggml/ggml.go
@@ -895,8 +895,8 @@ func (f GGML) FlashAttention() bool {
return slices.Contains([]string{
"gemma3",
"gptoss", "gpt-oss",
- "qwen3",
- "qwen3moe",
+ "qwen3", "qwen3moe",
+ "qwen3vl", "qwen3vlmoe",
}, f.KV().String("general.architecture"))
}