Grafana AI is an AI enhancement capability launched by Grafana Labs based on its open source observability stack, covering AI-driven alerts, natural language queries, automatic anomaly detection, and intelligent dashboard recommendations.

GrafanaAI

Core parameters and statistics

Grafana AI is a collection of AI capabilities built by Grafana Labs on the Grafana open source observability platform. Grafana is the industry's most popular open source dashboard platform. What's unique about its AI capabilities is that it is not a closed SaaS function, but is provided both as an open source plug-in and as a cloud service.

Projects	Public Information
Official positioning	The open observability platform with AI
AI capability form	LLM Plugin, AI alarm suggestions, natural language query AI anomaly detection, dashboard automatic suggestions
Core product stack	Grafana (dashboard), Loki (log), Tempo (tracing), Mimir (metric)
AI Plugin	Grafana LLM Plugin (connect OpenAI/Azure/Anthropic, etc.)
Deployment model	Grafana Cloud (SaaS), Grafana OSS (open source self-hosted)
Community size	20 million+ users GitHub 68,000+ stars
Company valuation	Grafana Labs valued at approximately $6 billion (unlisted)

Open source driven AI path: Grafana's AI strategy is different from Datadog and Dynatrace - Grafana does not have its own large-scale AI model, but allows users to choose to access external LLM (OpenAI, Anthropic, Azure OpenAI, local models, etc.) through the LLM Plugin. This means that the cost of using AI capabilities is entirely determined by the LLM selected by the user, and Grafana itself does not charge an AI premium.

The dividing line between open source and cloud services: AI alarms suggest that the open source version of LLM Plugin, natural language query and other functions requires self-configuration of LLM API keys and infrastructure; the Grafana Cloud version has built-in LLM connections and AI capabilities and does not require self-configuration. For a self-hosted team, the total cost of AI capabilities = self-hosted infrastructure cost + LLM API call fee.

User and market recognition

Grafana is the world's most widely used open source observability dashboard platform, with extremely high awareness and adoption in the open source community.

Open source community size: Grafana has 68,000+ stars and 12,000+ forks on GitHub. It is one of the open source projects with the largest Docker usage in the world. This means recruiting Grafana-related operations talent is easier than for proprietary platforms.

Enterprise customer base: Grafana Cloud serves 10,000+ enterprise customers, covering the Internet, finance, manufacturing and government industries. Grafana Labs’ ARR (annual recurring revenue) in 2025 is publicly reported to be close to $200 million.

Competitive Positioning: Grafana is uniquely positioned in the observability market - it is an "open source base" rather than a "full stack platform". Many Grafana users use both Datadog or Dynatrace as data sources, with Grafana used as a unified dashboard layer. This means that the value of Grafana AI is to enable open source stacks to have AI capabilities close to those of commercial platforms without having to migrate to closed platforms.

Cost advantage

Grafana AI’s cost structure is the most flexible in the observability market, as the primary cost of AI capabilities comes from LLM calls, not the Grafana license itself.

C client/individual: Grafana OSS is completely free. AI functions (LLM Plugin, natural language query) require self-configuration of LLM API keys. Individual developers can use the free quota of OpenAI API to experience AI functions, and the total cost is almost zero.

API/Developer: The Grafana API is completely free in the OSS version and included in the subscription plan in the Cloud version. API access is unlimited, but the LLM Plugin calling fee is paid directly by the user to the LLM provider (such as OpenAI, Anthropic).

Enterprise/Private: The Grafana Cloud Pro plan starts at $49/month (includes 3 users, a certain amount of metrics and log storage), and the Advanced/Enterprise plan is billed based on data volume and number of users. The biggest cost advantage for enterprises is that if they choose OSS self-hosting + on-premises LLM, they can avoid any platform license fees, which is an option that Datadog and Dynatrace cannot provide. Hidden costs: The maintenance manpower, storage costs (especially log storage), and API charges for LLM calls of a self-hosted Grafana + Loki + Tempo cluster may exceed expectations in large-scale usage scenarios.

Main functions

Grafana AI capabilities are delivered as open source plug-ins and Cloud functions:

• Grafana LLM Plugin: Open source plug-in that allows querying data using natural language in Grafana dashboards, generating dashboard descriptions and root cause analysis recommendations. Users need to configure an LLM API key (OpenAI, Azure OpenAI, Anthropic, Gemini are all acceptable), or they can access privately deployed LLM (such as the open source model run by Ollama). This "bring your own model" model allows data-sensitive teams to fully control the data flow for AI inference.
• AI Alarm Recommendation: Based on historical alarm data and indicator trends, AI automatically recommends thresholds and conditions for alarm rules. The value lies in reducing the trial and error process for operation and maintenance personnel to manually set each indicator threshold. For newly added indicators, AI can automatically give suggested thresholds based on historical data.
• Natural Language Query Dashboard: In the Grafana Explore interface, users can describe the data to be queried in natural language, and AI automatically generates corresponding PromQL/LogQL/TraceQL queries. This is useful for developers on the team who are not proficient in PromQL - "What was the 99th percentile latency of the payment service last week" can be automatically translated into the correct query statement.
• AI Anomaly Detection (Experimental): Automatic anomaly detection based on machine learning to identify abnormal patterns in time series indicators. Currently in an experimental stage, it is not as mature as Datadog Watchdog or Dynatrace Davis AI.
• Dashboard automatic suggestions: AI recommends relevant panel indicators and visualization methods based on the structure of data sources and historical query patterns, reducing the workload of building dashboards from scratch.

Model and version evolution

The relevant information has not been made public, please refer to the official real-time page.

Technical advantages

The technical advantages of Grafana AI come from the unique model of "open source ecosystem + plug-in architecture + user-selected LLM":

Open Source Plugin Architecture: The LLM Plugin is designed to decouple AI capabilities from the Grafana core. AI functions are not bound to the Grafana version. Even if users use an old version of Grafana OSS, they can still obtain AI capabilities by installing the latest version of LLM Plugin. This architecture is more flexible than closed platform AI integration.

Bring your own model (BYOM) mode: Grafana does not preset LLM providers, users can choose to connect to cloud LLM (OpenAI, Anthropic) or local private models (through Ollama or LLM API gateway). For regulated industries or data privacy-sensitive organizations, the local LLM path allows AI inference to be completed without leaving the intranet at all.

Data source agnostic: One of Grafana’s core strengths is its ability to connect to dozens of different types of data sources (Prometheus, Graphite, InfluxDB, Elasticsearch, CloudWatch, Azure Monitor, etc.). Grafana AI’s query capabilities cover query languages ​​for all of these data sources, rather than supporting just one.

How to use

Grafana provides two deployment paths: OSS self-hosting and Grafana Cloud. The entrance to use AI capabilities is unified in the Web UI:

How to use	Suitable for people	Features	Cost
Grafana Cloud	Teams who want to be operational-free	SaaS hosting, built-in LLM connection	Billed by subscription plan (Free/Pro/Advanced)
Grafana OSS self-hosting	Organizations with high data sovereignty requirements	Complete freedom, need to maintain the cluster by yourself	Infrastructure + operation and maintenance manpower
LLM Plugin (open source)	All users	After configuring the LLM API, it can be used in Explore	LLM API fees are borne by the user
REST API	Developer	Data query, panel management, alarm configuration	OSS free, Cloud according to plan

Typical usage process (self-hosted): Deploy Grafana OSS → Install LLM Plugin → Configure LLM API key (or local Ollama endpoint) → Use natural language query in Explore → Enable AI alert suggestions. For new users, it is recommended to start with the Grafana Cloud Free plan (14-day unlimited trial, which can be downgraded to the Free level later), and then evaluate the self-hosting needs after experiencing the AI ​​functions.

Product Pricing

Grafana’s pricing is known for its cost-effectiveness in the observability market, especially since the AI capabilities of the OSS self-hosted version are controlled by the users themselves.

• C-side/Personal: Grafana OSS completely free + free LLM Plugin. Individual developers only need to bear the cost of LLM API calls (such as OpenAI API pay-per-use), and the monthly cost can be controlled within a few dollars.
• API/Developer: API is unlimited in the OSS version. Grafana Cloud’s Free plan includes a basic API access limit, and the Pro plan starts at $49/month.
• Enterprise: Grafana Cloud Pro (starting at $49/month) and Advanced (business confirmation required) plans include AI capabilities. For self-hosted enterprises, the infrastructure cost depends on the amount of data (metric base, log throughput, tracking volume). It is recommended to use Grafana Cloud to estimate resource consumption before deciding whether to self-host. Hidden costs: The storage cost of self-hosted Grafana + Loki + Mimir may become a major expense in large-scale scenarios; and the cost of LLM API calls requires the establishment of usage tracking and budget control in large-scale usage scenarios.

Application scenarios

The application scenarios of Grafana AI cover multiple levels from daily queries to intelligent operation and maintenance:

• Natural language query for operation and maintenance data: For developers and managers who are not proficient in PromQL/LogQL, natural language query lowers the threshold for obtaining data from the monitoring platform. Developers can use everyday language to query data when troubleshooting problems, and do not need to ask an SRE to help write query statements.
• Automatic recommendation of alarm thresholds: When new services are launched or indicators change, the alarm thresholds automatically recommended by AI can reduce the workload of operation and maintenance personnel to manually adjust the thresholds. It is suitable for situations with a large index base and frequent changes.
• AI upgrade of the open source observability stack: Teams already using Grafana + Prometheus + Loki can access AI capabilities at low cost through the LLM Plugin without migrating to Datadog or Dynatrace. This is particularly attractive to organizations that already have an investment in Grafana.

Applicable people

Grafana AI’s open source positioning enables it to cover a wider user base than commercial platforms:

• DevOps and SRE Engineers: Teams that already have experience using Grafana OSS can gradually enhance the AI capabilities of the existing stack through LLM Plugin and AI alarm suggestions. No platform migration is required.
• Individual Developers and Open Source Contributors: Grafana OSS’s free AI capabilities allow individual developers to learn observability AI concepts without incurring licensing fees. Suitable for technology exploration and learning scenarios.
• Data Privacy Sensitive Enterprises: The path to self-hosted Grafana + on-premises LLM enables AI inference to be done completely offline and in context. This is especially important for regulated industries such as finance, healthcare, and government.

Not suitable for scenarios: Teams looking for "out-of-the-box full-stack observability" - Grafana AI's capabilities require self-configuration and assembly, which is not as convenient as the integrated solutions of Datadog or Dynatrace. In addition, if the team lacks Grafana operation and maintenance experience, the upfront investment and ongoing tuning costs of the self-hosted solution may be higher than that of the SaaS solution.

Summary and Outlook

The core competitiveness of Grafana AI lies in "open source base + self-contained model + zero license fee AI capabilities". It’s not the most AI-capable observability platform, but it’s the lowest-cost option with the most control. For organizations already using Grafana, the LLM Plugin provides a low-risk transition path from traditional dashboards to AI-driven observability.

Current limitations: AI anomaly detection is still in the experimental stage, and its maturity lags behind Datadog Watchdog and Dynatrace Davis AI; LLM Plugin's natural language query is not as accurate as the preset panel in complex aggregation scenarios; AI capabilities are scattered across multiple plug-ins, and the experience is not as smooth as the unified AI interface of commercial platforms.

Procurement/Adoption Risk Assessment: It is recommended to start with the Grafana Cloud Free plan or OSS + LLM Plugin to experience AI capabilities, and first use 2-4 weeks to verify the accuracy of natural language queries on your own data. If large-scale enterprises consider self-hosting solutions, they need to pre-evaluate the storage cost scaling trend of Loki/Mimir clusters and the cost prediction of LLM API calls at enterprise scale.

Version evolution of Grafana AI

Grafana releases major versions on a quarterly basis, and AI functions have been included in core capabilities since the 11.x series.

Grafana 11.x version evolution

• 11.0 (early 2025): LLM Plugin first released, experimental version supporting natural language → PromQL conversion.
• 11.1 (Mid 2025): AI alarm suggestion function introduced, LLM Plugin supports more LLM providers.
• 11.2 (Q3 2025): Natural language queries extend to LogQL and TraceQL.
• 11.3 (Late 2025): Natural Language Query Dashboard Beta, AI anomaly detection experimental improvements.
• 11.4 (early 2026): LLM Plugin GA, AI anomaly detection experimental version feature upgrade.
• 11.5 (Mid 2026): Enhanced AI alert suggestions, dashboard auto-suggestions Beta.

Self-hosted users are free to choose an upgrade window after a new version is released. Grafana Cloud users are automatically updated to the latest version.

Version Info

• Grafana 11.5 ：Released in the second quarter of 2026, new AI-driven alarm suggestions, Grafana LLM Plugin enhancements, and natural language query dashboard capabilities are upgraded. （2026-06-30）
• Grafana 11.4 ：Released in the first quarter of 2026, an experimental version of AI-driven anomaly detection, Grafana LLM Plugin GA, will be introduced. （2026-03-31）
• Grafana 11.3 ：Released in the fourth quarter of 2025, the natural language query dashboard function will be beta, and the AI ​​alarm suggestion function will be introduced. （2025-12-31）