Large language model performance matrix for Elastic Security
This page summarizes internal test results comparing large language models (LLMs) across Elastic Security AI chat and AI-powered feature use cases. The matrix tests each model across Agent Builder, Attack Discovery, and Automatic Migration. To learn more about these use cases, refer to AI-powered features.
Important
Higher scores indicate better performance, on a scale of 1 to 10. A score of 10 on a capability means the model met or exceeded all task-specific benchmarks for that capability.
Any model that scores 5 or below for a capability is not recommended for that task.
The matrix uses three top-line capability scores — Agent Builder, Attack Discovery, and Automatic Migration — that roll up into a single Overall Score. You can read the table top-down, from "how does this model perform across our AI features?" to "how good is it at the specific job I care about?"
- Overall Agent Builder Score is the average of the seven Agent Builder sub-capabilities (Alert Analysis, Entity Analytics, Threat Hunting, Detection Rules, Workflow Authoring, Triggering Workflows, and Multi-Step Executions). It summarizes how well a model handles agentic Security work end to end.
- Overall Score is the average of the Agent Builder, Attack Discovery, and Automatic Migration scores. It reflects how a model performs across the breadth of our AI features rather than any single workflow, and is the default sort for the tables below.
- Alert Analysis — Triage an alert, reach the correct disposition, pull related alerts, and enrich with threat intel.
- Entity Analytics — Investigate hosts and users using purpose-built entity lookups and risk context.
- Threat Hunting — Generate and run queries against process, file, and network telemetry to find specific hunt artifacts.
- Detection Rules — Author a working detection rule, grounded in research where requested.
- Workflow Authoring — Produce a valid, executable automation workflow (verified by actually creating, enabling, and running it).
- Triggering Workflows — Call the correct backed action for the task (for example, a hash lookup, an on-call schedule, or case creation).
- Multi-Step Executions — Chain several steps in the right order, carrying findings forward, without skipping or fabricating steps.
Models from third-party LLM providers.
| Model | Agent Builder: Alert Analysis | Agent Builder: Entity Analytics | Agent Builder: Threat Hunting | Agent Builder: Detection Rules | Agent Builder: Workflow Authoring | Agent Builder: Triggering Workflows | Agent Builder: Multi-Step Executions | Overall Agent Builder Score | Attack Discovery | Automatic Migration | Overall Score |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Anthropic Claude Opus 4.7 | 9.00 | 8.00 | 8.00 | 8.00 | 9.00 | 8.00 | 8.00 | 8.29 | 9.70 | 9.70 | 9.23 |
| Anthropic Claude Sonnet 4.5 | 9.00 | 7.00 | 7.00 | 7.00 | 9.00 | 8.00 | 8.00 | 7.86 | 9.10 | 10.00 | 8.99 |
| Anthropic Claude Sonnet 4.6 | 9.00 | 7.00 | 7.00 | 7.00 | 9.00 | 8.00 | 8.00 | 7.86 | 9.10 | 10.00 | 8.99 |
| OpenAI GPT-5.2 | 8.00 | 6.00 | 6.00 | 7.00 | 8.00 | 8.00 | 8.00 | 7.43 | 8.30 | 10.00 | 8.58 |
| Anthropic Claude Opus 4.6 | 9.00 | 8.00 | 7.00 | 8.00 | 9.00 | 8.00 | 8.00 | 8.14 | 7.50 | 10.00 | 8.55 |
| Google Gemini 2.5 Flash | 5.00 | 6.00 | 6.00 | 5.00 | 8.00 | 7.00 | 6.00 | 5.86 | 9.50 | 9.81 | 8.39 |
| Anthropic Claude Opus 4.5 | 9.00 | 7.00 | 7.00 | 8.00 | 9.00 | 8.00 | 8.00 | 8.00 | 9.20 | 7.30 | 8.17 |
| Anthropic Claude Haiku 4.5 | 6.00 | 7.00 | 7.00 | 5.00 | 9.00 | 8.00 | 8.00 | 6.71 | 6.50 | 10.00 | 7.74 |
| Google Gemini 3.1 Flash Lite | 7.00 | 7.00 | 7.00 | 7.00 | 8.00 | 8.00 | 8.00 | 7.57 | 3.20 | 9.90 | 6.89 |
| Google Gemini 3.0 Flash | 8.00 | 8.00 | 7.00 | 8.00 | 9.00 | 8.00 | 6.00 | 7.71 | 3.20 | 9.70 | 6.87 |
| OpenAI GPT-5.4 Mini | 7.00 | 7.00 | 7.00 | 7.00 | 8.00 | 8.00 | 6.00 | 7.14 | 3.50 | 9.80 | 6.81 |
| Google Gemini 3.5 Flash | 8.00 | 7.00 | 7.00 | 8.00 | 9.00 | 8.00 | 6.00 | 8.00 | 5.60 | 6.60 | 6.73 |
| Google Gemini 3.1 Pro (Preview) | 8.00 | 7.00 | 7.00 | 8.00 | 8.00 | 8.00 | 6.00 | 7.86 | 4.20 | 8.10 | 6.72 |
| OpenAI GPT-4.1 | 5.00 | 6.00 | 6.00 | 7.00 | 8.00 | 8.00 | 6.00 | 6.29 | 3.60 | 9.60 | 6.50 |
| OpenAI GPT-5.4 | 7.00 | 7.00 | 8.00 | 8.00 | 9.00 | 8.00 | 6.00 | 7.86 | 5.30 | 5.90 | 6.35 |
| OpenAI GPT-5.4 Nano | 5.00 | 3.00 | 5.00 | 7.00 | 8.00 | 8.00 | 6.00 | 5.57 | 1.00 | 9.90 | 5.49 |
| Google Gemini 2.5 Pro | 7.00 | 7.00 | 5.00 | 7.00 | 8.00 | 8.00 | 6.00 | 6.86 | 0.00 | 6.70 | 4.52 |
| Google Gemini 2.5 Flash Lite | 4.00 | 6.00 | 3.00 | 3.00 | 2.00 | 5.00 | 6.00 | 4.14 | 0.00 | 7.30 | 3.81 |
| OpenAI GPT-4.1 Mini | 6.00 | 3.00 | 6.00 | 5.00 | 9.00 | 7.00 | 3.00 | 5.57 | 4.20 | 0.00 | 3.26 |
Models you can deploy yourself.
| Model | Agent Builder: Alert Analysis | Agent Builder: Entity Analytics | Agent Builder: Threat Hunting | Agent Builder: Detection Rules | Agent Builder: Workflow Authoring | Agent Builder: Triggering Workflows | Agent Builder: Multi-Step Executions | Overall Agent Builder Score | Attack Discovery | Automatic Migration | Overall Score |
|---|---|---|---|---|---|---|---|---|---|---|---|
| OpenAI GPT-OSS 120B | 5.00 | 4.00 | 6.00 | 7.00 | 5.00 | 8.00 | 7.00 | 5.14 | 3.00 | 9.40 | 5.85 |
| Gemma 4 31B IT | 6.00 | 6.00 | 7.00 | 6.00 | 8.00 | 8.00 | 7.00 | 7.00 | 2.80 | 7.50 | 5.77 |
| DeepSeek V4 Pro | 5.00 | 6.00 | 6.00 | 6.00 | 9.00 | 8.00 | 7.00 | 5.86 | 8.30 | 3.10 | 5.75 |
| OpenAI GPT-OSS 20B | 4.00 | 5.00 | 5.00 | 7.00 | 5.00 | 8.00 | 4.00 | 5.43 | 2.60 | 4.00 | 4.01 |
| Qwen 3.6 27B | 5.00 | 7.00 | 7.00 | 8.00 | 9.00 | 8.00 | 6.00 | 7.14 | 0.00 | 4.10 | 3.75 |
| Kimi K2.6 | 0.00 | 6.00 | 0.00 | 0.00 | 9.00 | 6.00 | 7.00 | 4.00 | 0.00 | 3.10 | 2.37 |