目次
What is Ollama?
Ollama is a tool that allows you to run large language models (LLMs) like ChatGPT in your local environment.
For companies and individuals who value security and privacy, it is attracting attention as an AI chat platform that can be used without an internet connection.
Ollama installation instructions (Windows/macOS/Linux compatible)
Get the installer from the official Ollama website
- Website: Ollama official website
URL：https://ollama.com - Click “Download” or “Get started” to get the installer.
Performing the installation
- You can install it by simply starting the installer and following the instructions.
- No special settings are required, and you can start using it immediately after completion.
Launch Terminal (Command Line)
- Windows: PowerShell or Command Prompt
- macOS: Terminal
Specify where to save the model to be used in Ollama
Steps to add OLLAMA_MODELS in Windows (GUI)
The model is automatically downloaded from the internet and stored in the local environment.
The default download directory is:
Download destination path (default)
macOS / Linux
~/.ollama/modelsWindows (including WSL2 environment)
%USERPROFILE%\.ollama\modelsExample:
C:\Users\<あなたのユーザー名>\.ollama\modelsOptional: Specify where to save the model
OLLAMA_MODELS is an environment variable that is enabled by the user explicitly specifying .
If you don’t set this variable, Ollama will use the default save path (~/.ollama/models for example).
Open the environment variable settings screen
- Search in the Start menu
環境変数システム環境変数or enter
→ Click Edit Environment Variable or Edit System Environment Variable - When the System Properties screen opens,
→ At the bottom, click Environment Variable (N)…” Click
Add to User Environment Variables
- In the “User environment variable” column at the top, select “New (N)…” Click
- Enter the following: Field Entry
(e.g., variable nameOLLAMA_MODELS, variable valueC:\models, ← any destination folder) - Click OK to close
Reboot the relevant system
Command Prompt, PowerShell, Ollama and other related systems will restart
Make sure it’s set up correctly
Commands in PowerShell
$env:OLLAMA_MODELSCommands at the Command Prompt (cmd.exe)
echo %OLLAMA_MODELS% If the path you set (e.g., )C:\models is displayed, it is successful.
Notes
- Any folders
C:\modelsmust be created in advance. - With this setting, future Ollama models will be saved below this path.
Download the model (AI itself) to use in Ollama
Run any model (auto-download on first time)
ollama run モデル名How to choose a model
Let’s consider the balance between use, performance, PC specs, and size. I have organized it in an easy-to-understand manner below.
| viewpoint | substance | example |
|---|---|---|
| use | What do you want them to do? | Chatting, translating, code generation, image understanding, etc. |
| Model performance | How much accuracy and response quality do you want? | Focus on inference speed or accuracy |
| PC Performance | Do you have enough GPU, RAM, and storage space? | If the GPU is less than 8GB, a smaller model is recommended. |
Examples of the main models available
Models available in Ollama
https://ollama.com/search
There are many other models available, so choose the one that suits your application and performance.
| Model Name | Example Execution Command | size *Varies depending on the model variant | feature |
|---|---|---|---|
| deepseek-r1 | ollama run deepseek-r1 | 1GB～400GB | DeepSeek-R1 is an open inference model with performance close to leading models like O3 and Gemini 2.5 Pro. |
| gemma3 | ollama run gemma3 | 1GB~17GB | Gemma is a family of lightweight models offered by Google. The Gemma 3 model is multimodal (text and image processing), has a 128KB context window, and supports more than 140 languages. |
| qwen3 | ollama run qwen3 | 500MB～150GB | Qwen3 is the latest generation of large language models in the Qwen series, providing comprehensive high-density and mixed-expert (MoE) models |
| devstral | ollama run devstral | 14GB | Devstral: An Open Source Model for Coding Agents |
| llama4 | ollama run llama4 | 67GB～245GB | Meta’s latest collection of multimodal models. |
| llama3.3 | ollama run llama3.3 | 43GB | 70B model. Delivers the same performance as the Llama 3.1 405B model. |
| llama3.2 | ollama run llama3.2 | 1.3GB～2.0GB | It is miniaturized in 1B and 3B models. |
| llama3.1 | ollama run llama3.1 | 4.9GB～243GB | Available in 8B, 70B, and 405B parameter sizes. |
| llama3 | ollama run llama3 | 4.7GB | High-Performance, Open-Available LLMs |
| mistral | ollama run mistral | 4GB | Models Released by Mistral AI Lightweight and fast response performance |
| phi4 | ollama run phi4 | 9.1GB | Microsoft’s lightweight-oriented model |
Model download and run sample
ollama run llama3
Enter a anime in Send a message and the AI response will appear.
>>> あなたについて日本語で説明してください
私はLLaMAという名前のAIアシスタントです。私はMeta AIによって開発された巨大な言語モデルであり、人間と話すことができま
す。我々の主要な機能は、情報を提供し、質問に回答することです。また、自然発達的な会話にも対応します。
私は膨大なデータセット（インターネット、書籍など）で訓練されました。この訓練により、私は人間のような応答を生成するこ
とができます。我々の能力には以下のようなものがあります。
* 幅広いトピックに関する質問に回答
* 促子やトピックに基づいてテキストを生成
* 語言を翻訳する
* 長文を短縮したサマリーを生成
* 創作のアイデアや提案を提供
私は常に学習し、改善しています。私は完璧ではありませんが、私はお手伝いできるように設計されています！Continued use
Once you download the model, you can start it faster next time:
ollama run モデル名You can also check the downloaded models below:
ollama listList of Main Ollama CLI Commands
| command | use | example |
|---|---|---|
ollama | View a list of available subcommands | ollama |
ollama --version | Check the installed version | ollama --version |
ollama help | Get help | ollama help |
ollama list | View a list of locally saved models | ollama list |
ollama pull <モデル名> | Download the model (also automatically retrieved on run) | ollama pull モデル名 |
ollama run <モデル名> | Model Launch + Interaction Session | ollama run モデル名 |
ollama ps | View the list of models (processes) that are being started | ollama ps |
ollama stop <モデル名> | Stop running the specified model (ps after confirming with ) | ollama stop モデル名 |
ollama show <モデル名> | View model details (quantization format, license, parameters, etc.) | ollama show モデル名 |
ollama rm <モデル名> | Remove a model from local | ollama rm モデル名 |
ollama cp <元> <コピー名> | Copy the model (save it as an alias) | ollama cp モデル名 モデル名-copy |
ollama create <名前> -f <Modelfile> | Create a custom model using a modelfile | ollama create mymodel -f Modelfile |
ollama serve | Start the Ollama server. Used when using API integration or REST | ollama serve |
Benefits for Business Use
| item | substance |
|---|---|
| Offline use | No internet connection required, limited to in-house use |
| security | Secure handling of sensitive information without external transmission |
| Ease of Introduction | Execution environment can be set up in a few steps and is easy for IT to deploy |
| Flexible model selection | Switch between multiple models to optimize for your business needs |
Supplementary Information
- Ollama automatically leverages Docker and WSL2 on the backend, but users don’t need to be aware of it.
- If you are worried about command operation or want to operate with the Web UI, you can also link with Open WebUI.
Comments