Grafana Labs has released Grafana 12, the next major version of the popular open-source observability platform.
AI/ML Observability
New built-in dashboards and panels for monitoring machine learning pipelines:
- Model Performance Tracking — Monitor accuracy, drift, and inference latency in real-time
- GPU Monitoring — Native NVIDIA GPU utilization dashboards with DCGM exporter integration
- Experiment Tracking — Visualize ML experiment metrics from MLflow, Weights & Biases, and custom sources
Unified Alerting
The alerting system has been completely rebuilt:
- Single alert rule engine for all data sources
- Alert correlation and grouping to reduce noise
- SLO-based alerting with built-in error budget tracking
- Improved notification routing with on-call schedules
New Visualizations
- Flame Graphs — Native flame graph panel for profiling data (Pyroscope integration)
- Topology Maps — Service dependency visualization with health indicators
- Canvas Panel GA — Free-form custom visualization panels with data binding
