Skip to content

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

llm: Enable flash attention by default for gemma3

Jesse Grossc3c85aa06c1b
diff --git a/fs/ggml/ggml.go b/fs/ggml/ggml.go
index 920eff74a..fcb3d9fdb 100644
--- a/fs/ggml/ggml.go
+++ b/fs/ggml/ggml.go
@@ -893,6 +893,7 @@ func (f GGML) SupportsFlashAttention() bool {
// FlashAttention checks if the model should enable flash attention
func (f GGML) FlashAttention() bool {
return slices.Contains([]string{
+ "gemma3",
"gptoss", "gpt-oss",
"qwen3",
"qwen3moe",