Samsung Gauss

What is Samsung Gauss?

Samsung Gauss is a generative AI model developed in-house by Samsung Electronics. The second generation, Samsung Gauss2, handles language, code, and images together and is being rolled out in stages across internal operations and products.

• 1Samsung’s own in-house generative AI
• 2Three tiers: Compact, Balanced, and Supreme
• 3It is internal only, with no external sign-up

What sets it apart

These are the differences we found by running the same material through several tools in our training sessions.

Same job, different approach

Employees are pasting internal documents into external AI services to summarize them

Common uses

Who it’s for

Can we use it at work?

The questions that adoption leads and security teams ask first.

Please note: we could not confirm whether any new-generation model after Samsung Gauss2 has been officially announced. Nor has it been disclosed, feature by feature, which Galaxy device features run on this model.

10-minute quick start

1. 1

   Start by reading the Samsung Gauss2 announcement on the Samsung Electronics Newsroom. It lays out the model lineup and internal use cases in the company’s own words.

2. 2

   Browse the Samsung Developer Conference (SDC) Korea session list for talks on Code.i. They show most concretely how it is used internally.

3. 3

   Since you can’t try it yourself, run the same task on publicly available Korean models such as EXAONE or Mi:dm and compare.

4. 4

   Summarize the criteria for deciding whether your company needs its own model (data location, cost, owner) in a table and attach it to the approval document.

Copy-and-paste prompts

Replace only the [bracketed] parts, then paste it in.

You cannot enter this prompt into Samsung Gauss directly. Use it in a tool such as Claude or Gemini.

Outline a review report on whether our company should build its own AI model or buy an external one. Items: [where data is stored], [annual cost], [staff in charge], [how to roll back if it fails]. For each item, also list the questions we must confirm.

Use this prompt by pasting it into another tool to get a draft.

We are a company with [800] employees building an internal AI service. Design it separately for developers and office staff, and prioritize which features to open first for each. Also list the features that should not be opened first, and explain why.

A prompt to paste into another tool when you create training materials.

Write a 5-minute script explaining the difference between on-device AI and server-based AI to [non-technical office workers]. Compare the pros and cons of each across three dimensions: cost, speed, and security. Use just one analogy.

Draft with another tool, then have Legal review it.

Write a draft policy for employees using an internal-only AI. Use a table to separate what may be entered from what must not be entered, and describe the procedure if someone violates it. Leave the company name blank as [Company Name].

A prompt for pasting an official article into another tool.

From the article below, extract only the verifiable facts and create a one-page summary for executive reporting. Do not add anything that is not in the article, and mark which points the company itself stated and which are the reporter’s interpretation.

[Paste article text]

Using it with other tools

Linking two or three tools changes your work more than any single tool does.

Before you start

• “Using AI made by Samsung” usually refers to features on Galaxy devices. Don’t mix up the model name with device feature names.
• It is not a publicly available model, so do not fill in numbers in a performance comparison table. We have not measured any values ourselves.
• If you copy internal usage figures from a company announcement as they are, the numbers may be out of date by the time you use them.
• There is no way to inquire about adopting it. Don’t plan around a partnership or purchase.

Pricing

Samsung Gauss limitations

We don’t just list the positives. Here is what fell short when we used it.

• You can’t try it yourself

  It is the first model people name in a training session, yet there is nothing to show participants on screen. Explanations always end with a news quote, which leaves both the audience and the presenter unsatisfied.

• Updates are infrequent

  Official announcements come only when a new generation is released, so you can’t see what changes in between. Once you build materials, you have to keep re-checking that they are still accurate.

• Product features and model names get confused

  Galaxy’s AI features and this model’s name often get mixed up. The company has not said which feature runs on what, so when someone asks, you cannot give a definite answer.

Tools to compare

EXAONEA Korean model you can download and run→HyperCLOVA XBuilt into Naver’s services→WrtnA Korean AI you can use without logging in→Galaxy AICall recording, meeting minutes→A.XA Korean hyperscale model released in full→Mi:dmA small model that knows Korea→