Skip to content

Commit

Permalink
[8.x] Adds LLM matrix for OSS models (backport #6424) (#6459)
Browse files Browse the repository at this point in the history 
* Adds LLM matrix for OSS models (#6424)

* First draft

* Update llm-performance-matrix.asciidoc

* links from BYO page to performance matrix

* updates table format

* fixes broken ref

* inverts first matrix

* flips tables

* removes old tables

* incorporates last min edit from James

(cherry picked from commit 6d64b54)

# Conflicts:
#	docs/serverless/AI-for-security/connect-to-byo-llm.asciidoc
#	docs/serverless/AI-for-security/llm-performance-matrix.asciidoc

* Delete docs/serverless directory and its contents

---------

Co-authored-by: Benjamin Ironside Goldstein <[email protected]>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
  • Loading branch information
@mergify @benironside @github-actions
3 people authored Jan 22, 2025
1 parent 3bfaa14 commit 97bf833
Show file tree
Hide file tree
Showing 2 changed files with 33 additions and 11 deletions.
2 changes: 2 additions & 0 deletions docs/AI-for-security/connect-to-byo.asciidoc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,8 @@ image::images/lms-studio-arch-diagram.png[Architecture diagram for this guide]

NOTE: For testing, you can use alternatives to Nginx such as https://learn.microsoft.com/en-us/azure/developer/dev-tunnels/overview[Azure Dev Tunnels] or https://ngrok.com/[Ngrok], but using Nginx makes it easy to collect additional telemetry and monitor its status by using Elastic's native Nginx integration. While this example uses cloud infrastructure, it could also be replicated locally without an internet connection.

NOTE: For information about the performance of open-source models on tasks within {elastic-sec}, refer to the <<llm-performance-matrix,LLM performance matrix>>.

[discrete]
== Configure your reverse proxy

Expand Down
42 changes: 31 additions & 11 deletions docs/AI-for-security/llm-performance-matrix.asciidoc
View file
Original file line number Diff line number Diff line change
@@ -1,17 +1,37 @@
[[llm-performance-matrix]]
= Large language model performance matrix

This table describes the performance of various large language models (LLMs) for different use cases in {elastic-sec}, based on our internal testing. To learn more about these use cases, refer to <<attack-discovery, Attack discovery>> or <<security-assistant, AI Assistant>>.
This page describes the performance of various large language models (LLMs) for different use cases in {elastic-sec}, based on our internal testing. To learn more about these use cases, refer to <<attack-discovery, Attack discovery>> or <<security-assistant, AI Assistant>>.

[cols="1,1,1,1,1,1,1,1,1,1", options="header"]
NOTE: `Excellent` is the best rating, followed by `Great`, then by `Good`, and finally by `Poor`.

[discrete]
== Proprietary models
Models from third-party LLM providers.

[cols="1,1,1,1,1,1,1", options="header"]
|===
| *Feature* | | *Assistant - General* | *Assistant - {esql} generation* | *Assistant - Alert questions* | *Assistant - Knowledge retrieval* | *Attack Discovery*
| *Model* |*Claude 3: Opus* | Excellent | Excellent | Excellent | Good | Great
| |*Claude 3.5: Sonnet v2*| Excellent | Excellent | Excellent | Excellent | Great
| |*Claude 3.5: Sonnet* | Excellent| Excellent | Excellent | Excellent | Excellent
| |*Claude 3.5: Haiku* | Excellent| Excellent | Excellent | Excellent | Poor
| |*Claude 3: Haiku* | Excellent| Excellent | Excellent | Excellent | Poor
| |*GPT-4o* | Excellent| Excellent | Excellent | Excellent | Great
| |*GPT-4o-mini* | Excellent| Great | Great | Great | Poor
| |**Gemini 1.5 Pro 002** | Excellent| Excellent | Excellent | Excellent | Excellent
| |**Gemini 1.5 Flash 002**|Excellent| Poor | Good | Excellent | Poor
|===
| *Feature* | *Model* | | | | | | | |
| | *Claude 3: Opus*| *Claude 3.5: Sonnet v2* | *Claude 3.5: Sonnet* | *Claude 3.5: Haiku*| *Claude 3: Haiku* | *GPT-4o* | *GPT-4o-mini* | **Gemini 1.5 Pro 002** | **Gemini 1.5 Flash 002**
| *Assistant - General* | Excellent | Excellent | Excellent | Excellent | Excellent | Excellent | Excellent | Excellent | Excellent
| *Assistant - {esql} generation*| Excellent | Excellent | Excellent | Excellent | Excellent | Excellent | Great | Excellent | Poor
| *Assistant - Alert questions* | Excellent | Excellent | Excellent | Excellent | Excellent | Excellent | Great | Excellent | Good
| *Assistant - Knowledge retrieval* | Good | Excellent | Excellent | Excellent | Excellent | Excellent | Great | Excellent | Excellent
| *Attack Discovery* | Great | Great | Excellent | Poor | Poor | Great | Poor | Excellent | Poor

[discrete]
== Open-source models
Models you can <<connect-to-byo-llm, deploy yourself>>.

[cols="1,1,1,1,1,1,1", options="header"]
|===

NOTE: `Excellent` is the best rating, followed by `Great`, then by `Good`, and finally by `Poor`.
| *Feature* | | *Assistant - General* | *Assistant - {esql} generation* | *Assistant - Alert questions* | *Assistant - Knowledge retrieval* | *Attack Discovery*
| *Model* | *Mistral Nemo* | Good | Good | Great | Good | Poor
| | *LLama 3.2* | Good | Poor | Good | Poor | Poor
| | *LLama 3.1 405b* | Good | Great | Good | Good| Poor
| | *LLama 3.1 70b* | Good | Good | Poor | Poor | Poor
|===

0 comments on commit 97bf833

Please sign in to comment.