diff --git a/pkg/aikit2llb/inference/convert.go b/pkg/aikit2llb/inference/convert.go
index 7ff59588..8e146272 100644
--- a/pkg/aikit2llb/inference/convert.go
+++ b/pkg/aikit2llb/inference/convert.go
@@ -17,7 +17,7 @@ const (
 	distrolessBase = "gcr.io/distroless/cc-debian12:latest"
 
 	localAIRepo    = "https://github.com/mudler/LocalAI"
-	localAIVersion = "v2.13.0"
+	localAIVersion = "v2.14.0"
 	cudaVersion    = "12-3"
 )
 
diff --git a/pkg/aikit2llb/inference/exllama.go b/pkg/aikit2llb/inference/exllama.go
index b52e51ce..8e511098 100644
--- a/pkg/aikit2llb/inference/exllama.go
+++ b/pkg/aikit2llb/inference/exllama.go
@@ -26,7 +26,7 @@ func installExllama(c *config.InferenceConfig, s llb.State, merge llb.State) llb
 	s = cloneLocalAI(s, backend)
 
 	// clone exllama to localai exllama backend path and install python dependencies
-	s = s.Run(utils.Shf("git clone --depth 1 %[1]s --branch %[2]s /tmp/%[3]s && mv /tmp/%[3]s/* /tmp/localai/backend/python/%[3]s && rm -rf /tmp/%[3]s && cd /tmp/localai/backend/python/%[3]s && rm -rf .git && . $HOME/.cargo/env && uv pip install --no-build-isolation --requirement requirements-install.txt && EXLLAMA_NOCOMPILE= uv pip install --no-build-isolation && make protogen", exllamaRepo, exllamaTag, backend)).Root()
+	s = s.Run(utils.Shf("git clone --depth 1 %[1]s --branch %[2]s /tmp/%[3]s && mv /tmp/%[3]s/* /tmp/localai/backend/python/%[3]s && rm -rf /tmp/%[3]s && cd /tmp/localai/backend/python/%[3]s && rm -rf .git && . $HOME/.cargo/env && make exllama2", exllamaRepo, exllamaTag, backend)).Root()
 
 	// && pip3 install grpcio protobuf typing-extensions sympy mpmath setuptools numpy --break-system-packages && pip3 install -r /tmp/localai/backend/python/%[3]s/requirements.txt --break-system-packages