Grafana 11.5 released: AI alarm suggestions and natural language queries are in place, and the open source observable stack enters the "AI native" era

Grafana 11.5 was released on June 30, 2026. It is the second quarter version of Grafana Labs in 2026. The core updates focus on three AI main lines: AI-driven alarm suggestions, Grafana LLM Plugin enhancements, and natural language query dashboard capability upgrades. As an open source observable platform serving more than 20 million users, Grafana is upgrading AI from "plug-ins" to "platform native capabilities".

• AI-driven alarm suggestions: Based on historical data and indicator patterns, alarm rule setting suggestions are given to reduce the risk of alarm mismatches and missing alarms.
• Grafana LLM Plugin enhancement: LLM plug-in capabilities continue to be improved, allowing large models to access monitoring data context.
• Natural language query upgrade: Use natural language to describe query intentions to generate dashboards, lowering the threshold for data exploration.
• Continuation of AI anomaly detection: Taking over the experimental version of AI anomaly detection in 11.4, forming an AI closed loop of "detection-alarm-query".

Version background

Grafana Labs is the world's leading open source observability company, including Grafana, Loki, Tempo, Mimir and Grafana Cloud, serving 20 million+ users and 10,000+ enterprise customers. Its AI-based rhythm is clear: 11.3 (2025-12-31) introduces natural language query beta and AI alarm suggestions; 11.4 (2026-03-31) converts the Grafana LLM Plugin to regular (GA) and introduces an AI anomaly detection experimental version; 11.5 (2026-06-30) upgrades the above capabilities as a whole. The three versions constitute a complete advancement path of "Experiment-GA-Enhancement".

Highlights of this version

AI Alarm Suggestions

• Intelligent Rule Recommendation: AI analyzes the historical patterns of indicators and recommends reasonable alarm thresholds and rules to reduce manual trial and error.
• Reduce the cost of false alarms: More reasonable alarm configurations directly improve the "alarm fatigue" problem and allow operation and maintenance to focus on real problems.

Natural language query and LLM

• Natural Language Dashboard: Use natural language to describe query intentions, and AI generates corresponding queries and dashboards, lowering the threshold for novices to use.
• LLM Plugin enhancement: Call large models in the monitoring context to support natural language interactive analysis of logs/indicators.

AI Anomaly Detection

• Automatic detection of anomalies: Automatically identify fluctuations that deviate from the normal pattern based on indicator timing, exposing potential faults in advance.
• Linkage with alarms: Anomaly detection results directly enter the alarm link, forming an automated response closed loop.

Significance to the operation and maintenance team

From an industry perspective, Grafana's AI path is a benchmark for the open source observable track: it does not pursue "black box AI", but embeds AI capabilities into mature open source stacks in the form of plug-ins and modules, allowing teams to enable them on demand. For domestic operation and maintenance teams, Grafana's open source attributes mean that it can be self-hosted and controllable data, and its AI capabilities (natural language query, alarm suggestions) lower the threshold for use and maintenance of the monitoring system.

In comparison, Sentry focuses on application troubleshooting and error attribution, while Grafana focuses on indicator/log/tracking observation and visualization. The two are complementary rather than substitutes in terms of "AI observability".

Tips for getting started

• Self-hosted team: Experience natural language query and AI alarm suggestions from the OSS version, and evaluate improvements to the existing monitoring process.
• Cloud users: Quickly enable LLM Plugin and anomaly detection through Grafana Cloud, and verify AI capabilities with a low threshold.
• Evaluation Criteria: Measure the actual value of AI capabilities from the two dimensions of "alarm accuracy improvement" and "query analysis time reduction".

Directions worthy of attention in the future

1. The rhythm of turning AI anomaly detection into a positive one: when will the experimental version’s capabilities be GA, and how the false alarm rate will perform in real production.
2. Natural language query coverage depth: Whether it can cover complex multi-source queries determines its transition from "early adopter" to "main force".
3. The boundary between open source and commercial versions: The distribution of AI capabilities between OSS and Cloud affects the choice of self-hosted teams.