diff --git a/backend/routes/ollama_routes.py b/backend/routes/ollama_routes.py
index 50bfb18..822d79d 100644
--- a/backend/routes/ollama_routes.py
+++ b/backend/routes/ollama_routes.py
@@ -221,19 +221,41 @@ def get_gpu_status():
                         import subprocess
                         # Check if nvidia-smi works in ollama container
                         result = subprocess.run(
-                            ['docker', 'exec', 'munich-news-ollama', 'nvidia-smi', '--query-gpu=name', '--format=csv,noheader'],
+                            ['docker', 'exec', 'munich-news-ollama', 'nvidia-smi', '--query-gpu=name,memory.used,utilization.gpu', '--format=csv,noheader'],
                             capture_output=True,
                             text=True,
                             timeout=5
                         )
                         if result.returncode == 0 and result.stdout.strip():
+                            gpu_data = result.stdout.strip().split(',')
                             gpu_info['gpu_available'] = True
                             gpu_info['gpu_details'] = {
-                                'gpu_name': result.stdout.strip(),
+                                'gpu_name': gpu_data[0].strip() if len(gpu_data) > 0 else 'Unknown',
+                                'memory_used': gpu_data[1].strip() if len(gpu_data) > 1 else 'N/A',
+                                'utilization': gpu_data[2].strip() if len(gpu_data) > 2 else 'N/A',
                                 'note': 'GPU available but no model currently loaded'
                             }
-                    except:
-                        pass
+                            
+                            # Check Ollama logs for GPU usage evidence
+                            log_result = subprocess.run(
+                                ['docker', 'logs', '--tail', '50', 'munich-news-ollama'],
+                                capture_output=True,
+                                text=True,
+                                timeout=5
+                            )
+                            if log_result.returncode == 0:
+                                logs = log_result.stdout + log_result.stderr
+                                # Look for GPU offloading messages
+                                if 'offloaded' in logs.lower() and 'gpu' in logs.lower():
+                                    gpu_info['gpu_in_use'] = True
+                                    gpu_info['gpu_details']['note'] = 'GPU acceleration active (detected in logs)'
+                                    # Extract layer info if available
+                                    import re
+                                    match = re.search(r'offloaded (\d+)/(\d+) layers', logs, re.IGNORECASE)
+                                    if match:
+                                        gpu_info['gpu_details']['layers_offloaded'] = f"{match.group(1)}/{match.group(2)}"
+                    except Exception as e:
+                        gpu_info['debug_error'] = str(e)
                 
                 # Try to get system info
                 try: