diff --git a/pkg/inference/backends/llamacpp/llamacpp.go b/pkg/inference/backends/llamacpp/llamacpp.go
index db28aca..dd2a8d4 100644
--- a/pkg/inference/backends/llamacpp/llamacpp.go
+++ b/pkg/inference/backends/llamacpp/llamacpp.go
@@ -9,7 +9,7 @@ import (
 	"os/exec"
 	"path/filepath"
 	"runtime"
-	"slices"
+	"strconv"
 
 	"github.com/docker/model-runner/pkg/inference"
 	"github.com/docker/model-runner/pkg/inference/models"
@@ -76,7 +76,7 @@ func (l *llamaCpp) Install(ctx context.Context, httpClient *http.Client) error {
 	if runtime.GOOS == "linux" {
 		return errors.New("not implemented")
 	} else if (runtime.GOOS == "darwin" && runtime.GOARCH == "amd64") ||
-		(runtime.GOOS == "windows" && !slices.Contains([]string{"amd64", "arm64"}, runtime.GOARCH)) {
+		(runtime.GOOS == "windows" && !(runtime.GOARCH == "amd64" || runtime.GOARCH == "arm64")) {
 		return errors.New("platform not supported")
 	}
 
@@ -115,6 +115,11 @@ func (l *llamaCpp) Run(ctx context.Context, socket, model string, mode inference
 		return fmt.Errorf("failed to get model path: %w", err)
 	}
 
+	modelDesc, err := l.modelManager.GetModel(model)
+	if err != nil {
+		return fmt.Errorf("failed to get model: %w", err)
+	}
+
 	if err := os.RemoveAll(socket); err != nil {
 		l.log.Warnln("failed to remove socket file %s: %w", socket, err)
 		l.log.Warnln("llama.cpp may not be able to start")
@@ -124,10 +129,28 @@ func (l *llamaCpp) Run(ctx context.Context, socket, model string, mode inference
 	if l.updatedLlamaCpp {
 		binPath = l.updatedServerStoragePath
 	}
-	llamaCppArgs := []string{"--model", modelPath, "--jinja", "-ngl", "100"}
+	llamaCppArgs := []string{"--model", modelPath, "--jinja"}
 	if mode == inference.BackendModeEmbedding {
 		llamaCppArgs = append(llamaCppArgs, "--embeddings")
 	}
+	if runtime.GOOS == "windows" && runtime.GOARCH == "arm64" {
+		// Using a thread count equal to core count results in bad performance, and there seems to be little to no gain
+		// in going beyond core_count/2.
+		// TODO(p1-0tr): dig into why the defaults don't work well on windows/arm64
+		nThreads := min(2, runtime.NumCPU()/2)
+		llamaCppArgs = append(llamaCppArgs, "--threads", strconv.Itoa(nThreads))
+
+		modelConfig, err := modelDesc.Config()
+		if err != nil {
+			return fmt.Errorf("failed to get model config: %w", err)
+		}
+		// The Adreno OpenCL implementation currently only supports Q4_0
+		if modelConfig.Quantization == "Q4_0" {
+			llamaCppArgs = append(llamaCppArgs, "-ngl", "100")
+		}
+	} else {
+		llamaCppArgs = append(llamaCppArgs, "-ngl", "100")
+	}
 	llamaCppProcess := exec.CommandContext(
 		ctx,
 		filepath.Join(binPath, "com.docker.llama-server"),