Using Gemma 4 with profClaw for free local AI

[thegdsks]

Gemma 4 E4B dropped April 2026. Apache 2.0 license. Native tool calling. 128K context. Runs on a 16GB Mac via Ollama. Costs nothing.

We added support in v2.2.0. Here is the setup:

1. Update Ollama to 0.20+:

brew upgrade ollama

2. Pull Gemma 4:

ollama pull gemma4:e4b

3. Use it in profClaw:

profclaw chat -m gemma4:e4b 'What can you help me with?'

Or via API:

curl -X POST http://localhost:3000/api/chat/send \
  -H 'Content-Type: application/json' \
  -d '{"message":"hello","model":"gemma4:e4b"}'

Token usage: 16-24 input tokens per message (we optimized the system prompt). Tool calling works, the model calls web_search, exec, read_file correctly.

Also added Qwen 3 4B which is even better at tool calling if you want an alternative.

Anyone else running local models with profClaw? What is your setup?