Instructions to use rexium-ai/Ingot-2B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use rexium-ai/Ingot-2B 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 rexium-ai/Ingot-2B:Q6_K
# Run inference directly in the terminal:
llama cli -hf rexium-ai/Ingot-2B:Q6_K

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf rexium-ai/Ingot-2B:Q6_K
# Run inference directly in the terminal:
llama cli -hf rexium-ai/Ingot-2B:Q6_K

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 rexium-ai/Ingot-2B:Q6_K
# Run inference directly in the terminal:
./llama-cli -hf rexium-ai/Ingot-2B:Q6_K

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 rexium-ai/Ingot-2B:Q6_K
# Run inference directly in the terminal:
./build/bin/llama-cli -hf rexium-ai/Ingot-2B:Q6_K

Use Docker

docker model run hf.co/rexium-ai/Ingot-2B:Q6_K

LM Studio
Jan

vLLM

How to use rexium-ai/Ingot-2B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "rexium-ai/Ingot-2B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rexium-ai/Ingot-2B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/rexium-ai/Ingot-2B:Q6_K

Ollama
How to use rexium-ai/Ingot-2B with Ollama:
```
ollama run hf.co/rexium-ai/Ingot-2B:Q6_K
```

Unsloth Studio

How to use rexium-ai/Ingot-2B 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 rexium-ai/Ingot-2B 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 rexium-ai/Ingot-2B to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for rexium-ai/Ingot-2B to start chatting

How to use rexium-ai/Ingot-2B with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf rexium-ai/Ingot-2B:Q6_K

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": "rexium-ai/Ingot-2B:Q6_K"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

OpenClaw new

How to use rexium-ai/Ingot-2B with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf rexium-ai/Ingot-2B:Q6_K

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://127.0.0.1:8080/v1 \
  --custom-model-id "rexium-ai/Ingot-2B:Q6_K" \
  --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"

Docker Model Runner
How to use rexium-ai/Ingot-2B with Docker Model Runner:
```
docker model run hf.co/rexium-ai/Ingot-2B:Q6_K
```

Lemonade

How to use rexium-ai/Ingot-2B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull rexium-ai/Ingot-2B:Q6_K

Run and chat with the model

lemonade run user.Ingot-2B-Q6_K

List all available models

lemonade list

Hermes Agent

How to use rexium-ai/Ingot-2B with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf rexium-ai/Ingot-2B:Q6_K

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 rexium-ai/Ingot-2B:Q6_K

Run Hermes

hermes

Atomic Chat

Ingot-2B / README.md

rexium-ai

docs: language_bcp47 pt-PT (Hub YAML rules)

f46d805 verified 1 day ago

preview code

Raw

History Blame Contribute Delete

4.49 kB

	---
	license: apache-2.0
	base_model: Qwen/Qwen3.5-2B
	language:
	- pt
	- en
	language_bcp47:
	- pt-PT
	- en-US
	tags:
	- rexium
	- ingot
	- quantization
	- portuguese
	- gguf
	pipeline_tag: text-generation
	homepage: https://rexium.ai
	---

	<div align="center">

	# Ingot-2B

	### The compressed base from [Rexium](https://rexium.ai)

	[![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
	[![Hub](https://img.shields.io/badge/🤗-rexium--ai%2FIngot--2B-yellow.svg)](https://huggingface.co/rexium-ai/Ingot-2B)
	[![Access](https://img.shields.io/badge/Access-Gated%20(manual)-orange.svg)](https://huggingface.co/rexium-ai/Ingot-2B)

	~1.45 GiB GGUF · PT-PT + EN · One base → many specialists

	</div>

	---

	## Why Ingot

	In a forge, the ingot is the billet you cast once — correct alloy, ready to hammer into tools.

	Ingot-2B is that billet for Rexium: a small, compressed foundation so you can run many vertical specialists (`Ingot-2B-sport`, `Ingot-2B-fiscal`, …) without paying for a full-size model every time.

	We compete on fit — Portuguese (Portugal) + English for real products — not on being another generic tiny LLM.

	> Built on [`Qwen/Qwen3.5-2B`](https://huggingface.co/Qwen/Qwen3.5-2B) (Apache-2.0). We claim the compression and language-fit work on top, not the upstream pre-training.

	---

	## Highlights

	\| \| \|
	\|---\|---\|
	\| Ship artefact \| `Ingot-2B-Q6_K.gguf` ≈ 1.45 GiB (Q6_K) \|
	\| Languages \| PT-PT and EN (product focus) \|
	\| Runtime \| llama.cpp / GGUF-friendly stacks (CUDA cloud validated internally) \|
	\| Family \| `Ingot-2B` base → `Ingot-2B-<vertical>` specialists \|
	\| Access \| Gated (manual) — page is public; weights only after Rexium approval \|

	---

	## Model overview

	\| Characteristic \| Detail \|
	\| --- \| --- \|
	\| Base model \| [`Qwen/Qwen3.5-2B`](https://huggingface.co/Qwen/Qwen3.5-2B) (Apache-2.0) \|
	\| What we ship \| Merged language-adapted weights, exported as GGUF Q6_K \|
	\| Size on disk \| ≈ 1.45 GiB \|
	\| Intended role \| Base for further LoRA / specialist fine-tunes — not a finished vertical \|
	\| Org \| [`rexium-ai`](https://huggingface.co/rexium-ai) \|

	---

	## Quick start (llama.cpp)

	After your access request is approved:

	```bash
	# download (requires HF token with access)
	huggingface-cli download rexium-ai/Ingot-2B Ingot-2B-Q6_K.gguf --local-dir ./ingot

	./llama-server -m ./ingot/Ingot-2B-Q6_K.gguf -ngl 99 --port 8080 --jinja
	```

	Then call the OpenAI-compatible endpoint on `http://127.0.0.1:8080/v1`.

	For Qwen3.5 chat templates, keep thinking/reasoning off unless you intentionally want chain-of-thought (same family behaviour as upstream Qwen3.5).

	---

	## The Ingot family

	\| Name \| Role \|
	\| --- \| --- \|
	\| Ingot-2B \| Compressed bilingual base (this card) \|
	\| Ingot-2B-sport \| Specialist forged for PeakRaptor / sports science (when published) \|
	\| Ingot-2B-\* \| Future verticals (fiscal, …) — only when they exist \|

	One name, many tools. We do not invent empty SKUs on this card.

	---

	## Intended use

	- Embedding a PT-PT/EN capable small model in products and cloud GPU inference
	- Starting point for private specialists (domain LoRA) without training a 7B+ from scratch
	- Evaluation and demos under Rexium’s gated distribution

	Not intended as: a drop-in replacement for large frontier models; a guarantee of native-level European Portuguese; or an open dump of partner IP.

	---

	## Limitations (honest)

	- Language quality is measured on Rexium’s internal rubric; treat PT-PT claims as strong for the size class, not “problem solved”.
	- Cloud acceptance used a relative Q6_K vs F16 check on the same CUDA runtime — useful for deployment, not a public leaderboard score.
	- Evaluation sets are sized for internal decisions; enterprise claims need a larger contested protocol.
	- Specialists trained on partner data stay private.

	---

	## Access

	This repository is gated with manual approval.

	1. Open [rexium-ai/Ingot-2B](https://huggingface.co/rexium-ai/Ingot-2B)
	2. Request access and say briefly what you’re building
	3. After approval, download `Ingot-2B-Q6_K.gguf`

	Unauthenticated downloads are rejected.

	---

	## Links

	- Product: [rexium.ai](https://rexium.ai)
	- Org: [huggingface.co/rexium-ai](https://huggingface.co/rexium-ai)
	- Upstream: [Qwen3.5-2B](https://huggingface.co/Qwen/Qwen3.5-2B)