diff --git a/backend/internal/client/client_test.go b/backend/internal/client/client_test.go
index b354f0f..29c4655 100644
--- a/backend/internal/client/client_test.go
+++ b/backend/internal/client/client_test.go
@@ -10,7 +10,7 @@ import (
 	"reichard.io/aethera/internal/store"
 )
 
-const model = "vllm-qwen3.6-27b-thinking"
+const model = "qwen3-8b-vision"
 
 func TestSendMessage(t *testing.T) {
 	t.Skip("requires live LLM API - run manually with: go test -run TestSendMessage ./internal/client/")
@@ -94,7 +94,10 @@ func TestSummarizeChat(t *testing.T) {
 }
 
 func TestSendMessageWithImage(t *testing.T) {
-	t.Skip("requires live LLM API - run manually with: go test -run TestSendMessageWithImage ./internal/client/")
+	// Skip: qwen3-8b-vision on LlamaSwap returns 502 for vision requests.
+	// Run manually when a stable vision backend is available:
+	//   go test -run TestSendMessageWithImage ./internal/client/
+	t.Skip("requires stable vision backend - LlamaSwap qwen3-8b-vision returns 502 for image inputs")
 
 	// Initialize Client
 	baseURL, err := url.Parse("https://llm-api.va.reichard.io/v1")
@@ -107,14 +110,15 @@ func TestSendMessageWithImage(t *testing.T) {
 	ctx, cancel := context.WithTimeout(context.Background(), 120*time.Second)
 	defer cancel()
 
+	// Build Data URL from embedded test image (10x10 red PNG)
+	const dataURL = "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAoAAAAKCAYAAACNMs+9AAAADklEQVR42mNk+P/k/wEAn6gE9QxN+7QAAAAASUVORK5CYII="
+
 	// Generate Text Stream
 	_, err = client.SendMessage(ctx, []*store.Message{{
 		Role:    "user",
 		Content: "What is in this image?",
-		Images: []string{
-			"https://llm-api.va.reichard.io/v1/images/test.png",
-		},
-	}}, "vllm-qwen3-8b-vision", func(mc *MessageChunk) error {
+		Images:  []string{dataURL},
+	}}, model, func(mc *MessageChunk) error {
 		if mc.Message != nil {
 			t.Logf("Received: %s", *mc.Message)
 		}