How to use from
OpenClaw
Start the llama.cpp server
# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf OpenTARS/Gargantua-Pro:Q4_K_M
Configure OpenClaw
# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:9931/v1 \
  --custom-model-id "OpenTARS/Gargantua-Pro:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Quick Links

Gargantua Pro (12B)

Gargantua Pro is the big model of the openTARS assistant, built for longer tasks: plan the steps, call the tools, check the result and keep going until the job is done. It uses the same 19 tools as Gargantua 2, on a larger base.

  • Base: Gemma 4 12B (Apache-2.0), fine-tuned by the openTARS project on its own tool-use data.
  • File: gargantua-pro-gemma4-12b-Q4_K_M.gguf (7.4 GB, text only).
  • Runs with: llama.cpp and Ollama (the gemma4 renderer and parser are built in). It needs a graphics card with about 8 GB, or plenty of RAM on the CPU.

Use

printf 'FROM ./gargantua-pro-gemma4-12b-Q4_K_M.gguf\nRENDERER gemma4\nPARSER gemma4\nPARAMETER temperature 0.1\n' > Modelfile
ollama create gargantua-pro -f Modelfile
ollama run gargantua-pro

It calls tools through the standard Ollama tool-calling API; the openTARS app sends the 19 tools for you.

Status

First release. We have not published benchmark numbers for it yet.

Follow the project at project-opentars.github.io/OpenTARS.

Downloads last month
93
GGUF
Model size
12B params
Architecture
gemma4
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support