Instructions to use everm4iva/archi2-2-9b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use everm4iva/archi2-2-9b with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="everm4iva/archi2-2-9b",
	filename="archi2-_q8.0_8b.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use everm4iva/archi2-2-9b with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf everm4iva/archi2-2-9b
# Run inference directly in the terminal:
llama-cli -hf everm4iva/archi2-2-9b

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf everm4iva/archi2-2-9b
# Run inference directly in the terminal:
llama-cli -hf everm4iva/archi2-2-9b

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 everm4iva/archi2-2-9b
# Run inference directly in the terminal:
./llama-cli -hf everm4iva/archi2-2-9b

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 everm4iva/archi2-2-9b
# Run inference directly in the terminal:
./build/bin/llama-cli -hf everm4iva/archi2-2-9b

Use Docker

docker model run hf.co/everm4iva/archi2-2-9b

LM Studio
Jan

vLLM

How to use everm4iva/archi2-2-9b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "everm4iva/archi2-2-9b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "everm4iva/archi2-2-9b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/everm4iva/archi2-2-9b

Ollama
How to use everm4iva/archi2-2-9b with Ollama:
```
ollama run hf.co/everm4iva/archi2-2-9b
```

Unsloth Studio

How to use everm4iva/archi2-2-9b 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 everm4iva/archi2-2-9b 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 everm4iva/archi2-2-9b to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for everm4iva/archi2-2-9b to start chatting

Docker Model Runner
How to use everm4iva/archi2-2-9b with Docker Model Runner:
```
docker model run hf.co/everm4iva/archi2-2-9b
```

Lemonade

How to use everm4iva/archi2-2-9b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull everm4iva/archi2-2-9b

Run and chat with the model

lemonade run user.archi2-2-9b-{{QUANT_TAG}}

List all available models

lemonade list

archi2 - A Logic-Optimized Reasoning Model

archi2 is an 9B parameter language model fine-tuned on top of Ministral-8B, trained on a limited and curated dataset up to May 2024.

It is optimized for raw logical reasoning, structured brainstorming, and high-quality human language expressiveness.

Model Summary

Property	Value
Base Model	Ministral-3B
Parameters	9B
Context Window	128,000 tokens
Training Data Cutoff	May 2024
Vision	✅ Images (no video/audio)
Function Calling	✅
License	PFF 1.0

Feel free to install. GGUF avaliable!

Intended Use

archi2 is a general-purpose reasoning model designed for:

Raw logic & logical deduction
Structured problem-solving and brainstorming
Expressive, nuanced language generation
Processing large documents and long-context tasks
Image understanding
Search and retrieval with context and filtering (truth and relevance)

Out of Scope

Real-time API or database integration (model is not designed for tool-augmented pipelines)
Audio or video understanding
Tasks requiring knowledge past May 2024
Advanced search and retrieval that requires up-to-date information or dynamic data sources

Persona

archi2 has developed a consistent internal persona through fine-tuning: curious, logic-first, and rigorously neutral. It approaches questions empirically, avoids dogma and "nonsense" contexts. It is expressive without being emotive, and precise without being cold.

It can be rude or blunt when the situation calls for it, but generally prefers clear and direct communication. It is not designed to be "friendly" or "polite" in a traditional sense, but rather to be an effective and efficient reasoning partner.

Recommended Inference Parameters

Parameter	Recommended Value
`temperature`	≤ 0.7 (less than)
`top_p`	≤ 0.78 (less than)
`frequency_penalty`	~0.6

Higher temperature or top_p values may introduce inconsistency in logical outputs. The defaults above strike a balance between creativity and coherence.

Function Calling

archi2 supports structured function calling. Pass tool definitions in the standard format and the model will respond with appropriate tool invocations when relevant.

Training Details

Base model: Ministral-3B
Fine-tuning data: Large-scale curated dataset spanning logic, reasoning, debate, science, philosophy, linguistics, and general knowledge — up to May 2024
Optimization focus: Logical coherence, reasoning depth, expressive language generation

Limitations

Knowledge cutoff is May 2024; it will not know about events after this date
Not suited for real-time or database-connected deployments in its current form
No audio or video modality support
Like all LLMs, it can produce plausible-sounding but incorrect outputs — always verify critical reasoning chains

License

This model is released under the PURE FREEDOM FOREVER (PFF 1.0) license.
See the LICENSE file for full terms.

Downloads last month: 158

GGUF

Model size

9B params

Architecture

mistral3

Hardware compatibility

We're not able to determine the quantization variants.

View all variants