Instructions to use IDEAHQ/ava-nautilus with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use IDEAHQ/ava-nautilus with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="IDEAHQ/ava-nautilus",
	filename="gguf/nemotron-3-nano-4b/NVIDIA-Nemotron3-Nano-4B-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use IDEAHQ/ava-nautilus with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf IDEAHQ/ava-nautilus:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf IDEAHQ/ava-nautilus:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf IDEAHQ/ava-nautilus:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf IDEAHQ/ava-nautilus:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf IDEAHQ/ava-nautilus:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf IDEAHQ/ava-nautilus:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf IDEAHQ/ava-nautilus:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf IDEAHQ/ava-nautilus:Q4_K_M

Use Docker

docker model run hf.co/IDEAHQ/ava-nautilus:Q4_K_M

LM Studio
Jan

vLLM

How to use IDEAHQ/ava-nautilus with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "IDEAHQ/ava-nautilus"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "IDEAHQ/ava-nautilus",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/IDEAHQ/ava-nautilus:Q4_K_M

Ollama
How to use IDEAHQ/ava-nautilus with Ollama:
```
ollama run hf.co/IDEAHQ/ava-nautilus:Q4_K_M
```

Unsloth Studio

How to use IDEAHQ/ava-nautilus with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for IDEAHQ/ava-nautilus to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for IDEAHQ/ava-nautilus to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for IDEAHQ/ava-nautilus to start chatting

How to use IDEAHQ/ava-nautilus with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IDEAHQ/ava-nautilus:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "IDEAHQ/ava-nautilus:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use IDEAHQ/ava-nautilus with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IDEAHQ/ava-nautilus:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default IDEAHQ/ava-nautilus:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use IDEAHQ/ava-nautilus with Docker Model Runner:
```
docker model run hf.co/IDEAHQ/ava-nautilus:Q4_K_M
```

Lemonade

How to use IDEAHQ/ava-nautilus with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull IDEAHQ/ava-nautilus:Q4_K_M

Run and chat with the model

lemonade run user.ava-nautilus-Q4_K_M

List all available models

lemonade list

Tali Nautilus

Tali Nautilus is the reasoning-focused on-device LLM family for TaliOS.

What it does

Where Tali Storm handles general dialogue and routine command interpretation, Nautilus is the model TaliOS reaches for when a user request requires multi-step reasoning or looking at images:

Multi-step planning — "find every email from the legal team about contract X this quarter, and summarize the open items" requires Nautilus to plan, dispatch sub-actions through TaliOS, and reason over the results.
Vision-language — "what is highlighted on the screen and what should I do next?" requires the VL variant to read pixels, not just accessibility metadata.

Where it sits in TaliOS

Nautilus runs only when Storm (the cheaper general-purpose LLM) signals it cannot complete the task. The TaliOS runtime escalates from Storm → Nautilus on a best-effort heuristic; the user never selects a model manually.

User speech → Tali STT → text → Tali NLU → vector match
                                     │
                                     ▼ (NLU miss)
                               Tali Storm → action / reply
                                     │
                                     ▼ (needs reasoning or vision)
                               Tali Nautilus → action / reply

Variants

Tali ID	Active Params	Architecture	Target
TALI-NAUTILUS-4B	4B	Dense	Phone
TALI-NAUTILUS-9B	9B	Dense	Desktop / tablet
TALI-NAUTILUS-30B-A3B	3B active	Mixture-of-Experts	Phone (flagship)
TALI-NAUTILUS-120B-A12B	12B active	Mixture-of-Experts	Desktop
TALI-NAUTILUS-CASCADE	3B active	MoE, reasoning-tuned	Phone (deep reasoning)
TALI-NAUTILUS-VL-8B	8B	Vision-Language	Vision tasks

Mixture-of-Experts (MoE) variants only load "active" parameters for any given token; memory cost is proportional to active params, not total.

Vision-Language (VL) accepts an image as part of its input — used for screen-content understanding and document scenes.

Quantization: Q4_K_M for on-device targets; full-precision weights retained for desktop.

File format

Shipped weights are wrapped in AON — Tali's encrypted, signed asset container. The .aon extension is the only format the OS or external tooling sees.

License

Proprietary — Intelligent Devices LLC.

Downloads last month: -

GGUF

Model size

4B params

Architecture

nemotron_h

Hardware compatibility

4-bit