--- title: "How to Set Up Gemma 4 with OpenClaw Using Ollama (2026 Guide)" description: "Run Google's Gemma 4 locally with Ollama and use it as your OpenClaw coding agent. Step-by-step Mac setup with copy-paste configs." date: 2026-04-04 updatedDate: 2026-04-27 location: San Francisco, CA – Apr 4th, 2026 image: /images/gemma-4-ollama-openclaw-setup-hero.jpg keywords: "gemma 4 openclaw, ollama gemma 4, openclaw local model, gemma 4 ollama setup, run gemma 4 locally, openclaw ollama guide" faq: - question: "Can I use Gemma 4 with OpenClaw?" answer: "Yes. Install Ollama, pull the gemma4 model, then add Ollama as a provider in ~/.openclaw/openclaw.json pointing to localhost:11434. OpenClaw treats Gemma 4 the same as any cloud model for tool calling and agentic workflows." - question: "How much memory does Gemma 4 need to run locally?" answer: "The default Gemma 4 8B model uses about 9.6GB when loaded. A Mac mini or MacBook with 16GB unified memory handles it comfortably. The 26B variant needs 24GB+ and will cause swapping on most consumer machines." - question: "Is Gemma 4 good enough for coding with OpenClaw?" answer: "Gemma 4 8B handles code reading, simple edits, boilerplate generation, and file operations well. For complex multi-file refactors or hard debugging, pair it with a cloud model through Haimaker's API for a hybrid local-plus-cloud setup." --- Google released Gemma 4 on April 2, 2026, and it runs on Ollama out of the box. If you're using OpenClaw as your coding agent, you can point it at a local Gemma 4 instance and skip the API bills for routine work. Recent developer threads and tutorials are all circling the same thing: people want a local coding agent that is cheap, private, and not miserable to set up. Gemma 4 + Ollama is the cleanest path right now, as long as you keep expectations reasonable. ## Gemma 4 Ollama quick commands For the impatient version: ```bash brew install --cask ollama-app open -a Ollama ollama pull gemma4 ollama run gemma4:latest "Write a tiny TypeScript function" ``` Then point OpenClaw at the local Ollama server: ```json5 { models: { providers: { ollama: { baseUrl: "http://localhost:11434/v1", apiKey: "ollama", api: "openai-completions", models: [{ id: "gemma4:latest", name: "Gemma 4" }] } } } } ``` That is the core setup. The rest of this guide covers the details that save you from annoying half-working states. Here's the full setup on a Mac with Apple Silicon: install Ollama, pull Gemma 4, and wire it into OpenClaw. ## What you need - Mac with Apple Silicon (M1/M2/M3/M4/M5) and at least 16GB unified memory - macOS with Homebrew installed - OpenClaw installed (`npm install -g openclaw`) The 8B default model uses about 9.6GB when loaded, leaving enough headroom on a 16GB machine. If you have 24GB or more, you'll barely notice it running. ## Step 1: Install Ollama Install the Ollama macOS app via Homebrew: ```bash brew install --cask ollama-app ``` This gives you `Ollama.app` in `/Applications/` and the `ollama` CLI at `/opt/homebrew/bin/ollama`. ## Step 2: Start Ollama ```bash open -a Ollama ``` The Ollama icon appears in the menu bar. Give it a few seconds to initialize, then verify: ```bash ollama list ``` ## Step 3: Pull Gemma 4 ```bash ollama pull gemma4 ``` This downloads roughly 9.6GB. Once it finishes, confirm the model is available: ```bash ollama list # NAME ID SIZE MODIFIED # gemma4:latest ... 9.6 GB ... ``` A quick sanity check: ```bash ollama run gemma4:latest "Hello, what model are you?" ``` Verify GPU acceleration is working: ```bash ollama ps # Should show CPU/GPU split, e.g. 14%/86% CPU/GPU ``` On Apple Silicon, Ollama v0.19+ automatically uses Apple's MLX framework for faster inference. No extra configuration needed. ## Step 4: Configure OpenClaw to use Gemma 4 The fastest way is the onboarding wizard: ```bash openclaw onboard --auth-choice ollama ``` Or add Ollama manually in `~/.openclaw/openclaw.json`: ```json5 { models: { providers: { ollama: { baseUrl: "http://localhost:11434/v1", api: "openai-completions", models: [ { id: "gemma4:latest", name: "Gemma 4 8B", reasoning: false, contextWindow: 131072, maxTokens: 8192 } ] } } }, agents: { defaults: { model: { primary: "ollama/gemma4:latest" }, models: { "ollama/gemma4:latest": { alias: "gemma4" } } } } } ``` Switch to Gemma 4 in OpenClaw: ``` /model gemma4 ``` ## Step 5: Keep Gemma 4 loaded and ready By default, Ollama unloads models after 5 minutes of inactivity. That means cold starts every time you come back from a coffee break. Fix it by setting the keep-alive to indefinite: ```bash launchctl setenv OLLAMA_KEEP_ALIVE "-1" ``` Then restart Ollama. To persist across reboots, add this to your `~/.zshrc`: ```bash export OLLAMA_KEEP_ALIVE="-1" ``` You can also set Ollama to launch at login: click the Ollama menu bar icon and enable **Launch at Login**. #### Auto-preload on startup Create a launch agent that warms the model after each reboot: ```bash cat << 'EOF' > ~/Library/LaunchAgents/com.ollama.preload-gemma4.plist Label com.ollama.preload-gemma4 ProgramArguments /opt/homebrew/bin/ollama run gemma4:latest RunAtLoad StartInterval 300 StandardOutPath /tmp/ollama-preload.log StandardErrorPath /tmp/ollama-preload.log EOF launchctl load ~/Library/LaunchAgents/com.ollama.preload-gemma4.plist ``` This sends an empty prompt every 5 minutes, keeping Gemma 4 warm in memory. ## What Gemma 4 handles well in OpenClaw After a few sessions with Gemma 4 running locally in OpenClaw, here's what it's actually good at: - **Reading and summarizing code.** Ask it to explain what a function does and you get a solid answer. It handles navigating unfamiliar codebases reasonably well. - **Boilerplate and scaffolding.** Config files, CRUD operations, test templates, simple React components. It writes functional code on the first try for common patterns. - **File operations.** Listing files, searching for patterns, renaming variables. Mechanical work that doesn't need deep reasoning. - **Quick edits.** Single-file changes, fixing typos, updating imports, adding a new field to a struct. ## Where it falls short - **Multi-file refactors.** Anything touching 5+ files gets unreliable. The model loses track of changes across files. - **Complex debugging.** If a bug spans multiple abstraction layers, Gemma 4 8B tends to suggest surface-level fixes. This is where bigger models earn their keep. - **Long context.** While Gemma 4 supports large context windows on paper, inference quality degrades on consumer hardware past 32K tokens. Keep your context window config realistic. ## Go hybrid: local Gemma 4 + cloud models through Haimaker Most people end up using Gemma 4 for routine work and sending harder tasks to cloud models. [Haimaker](https://haimaker.ai) makes this easy — one API key gets you access to Claude Opus, GPT-5, Gemini Pro, and others. Add it alongside Ollama in your OpenClaw config: ```json5 { models: { providers: { ollama: { baseUrl: "http://localhost:11434/v1", api: "openai-completions", models: [ { id: "gemma4:latest", name: "Gemma 4 8B", reasoning: false, contextWindow: 131072, maxTokens: 8192 } ] }, haimaker: { baseUrl: "https://api.haimaker.ai/v1", apiKey: "YOUR_HAIMAKER_API_KEY", api: "openai-completions", models: [ { id: "anthropic/claude-sonnet-4-6", name: "Claude Sonnet 4.6", reasoning: true, contextWindow: 200000, maxTokens: 16384 } ] } } }, agents: { defaults: { model: { primary: "ollama/gemma4:latest", thinking: "haimaker/anthropic/claude-sonnet-4-6" } } } } ``` Gemma 4 handles file reads, simple edits, and boilerplate — probably 60-70% of a typical session. Sonnet picks up the debugging and multi-file work. Your API bill drops to a few dollars a day instead of $20+. Switch manually when you hit something hard: ``` /model sonnet ``` Or use [Haimaker's auto-router](/blog/openclaw-auto-router-setup/) to detect task complexity and route automatically. Sign up at [haimaker.ai](https://haimaker.ai) to get your API key and browse the full [model catalog](https://haimaker.ai/models). GET YOUR HAIMAKER API KEY ## Troubleshooting **Model loads slowly or crashes.** You're probably running low on memory. Check what else is using your unified memory — close browser tabs running heavy WebGL or video. On a 16GB machine, Gemma 4 8B should load fine, but competing processes can push you into swap. **Tool calls fail.** Set `"reasoning": false` in your model config. Gemma 4 handles tool calling, but reasoning mode can cause formatting issues with OpenClaw's expected tool-call format. **Slow generation speed.** On Apple Silicon with Ollama v0.19+, you should see decent speeds thanks to the MLX backend. If you're getting unexpectedly slow output, make sure you're on the latest Ollama version — older versions don't use MLX. **Context window errors.** Set `contextWindow` to 131072 (128K) if you have 24GB+ memory. On 16GB, use 32768 to avoid quality degradation under memory pressure. ## Useful Ollama commands | Command | Description | |---|---| | `ollama list` | List downloaded models | | `ollama ps` | Show running models and memory usage | | `ollama run gemma4:latest` | Interactive chat | | `ollama stop gemma4:latest` | Unload model from memory | | `ollama pull gemma4:latest` | Update to latest version | | `ollama rm gemma4:latest` | Delete model | --- *For more local model options, see [best Ollama models for OpenClaw](/blog/best-local-models-for-openclaw). For a full comparison across all models, see [best models for OpenClaw](/blog/best-models-for-openclaw/). For cloud model pricing, check [cheapest models for OpenClaw](/blog/cheapest-models-openclaws).*