---
language:
- en
license: cc-by-nc-4.0
library_name: transformers
base_model: mistralai/Ministral-8B-Instruct-2410
base_model_relation: finetune
tags:
- conversational
- assistant
- fine-tuned
- lora
- collaborative
- vanta-research
- conversational-ai
- chat
- warm
- friendly-ai
- persona
- personality
- alignment
model-index:
- name: atom-v1-8b-preview
  results: []
---

<div align="center">

![vanta_trimmed](https://cdn-uploads.huggingface.co/production/uploads/686c460ba3fc457ad14ab6f8/hcGtMtCIizEZG_OuCvfac.png)
  
  <h1>VANTA Research</h1>
    
  <p><strong>Independent AI safety research lab specializing in cognitive fit, alignment, and human-AI collaboration</strong></p>
  
  <p>
    <a href="https://unmodeledtyler.com"><img src="https://img.shields.io/badge/Website-unmodeledtyler.com-yellow" alt="Website"/></a>
    <a href="https://x.com/vanta_research"><img src="https://img.shields.io/badge/@vanta_research-1DA1F2?logo=x" alt="X"/></a>
    <a href="https://github.com/vanta-research"><img src="https://img.shields.io/badge/GitHub-vanta--research-181717?logo=github" alt="GitHub"/></a>
  </p>
</div>

---

# Atom v1 8B Preview

Atom v1 8B Preview is a fine-tuned conversational AI model designed for collaborative problem-solving and thoughtful dialogue. Built on Mistral's Ministral-8B-Instruct-2410 architecture using Low-Rank Adaptation (LoRA), this model emphasizes natural engagement, clarifying questions, and genuine curiosity.

## Quick Start

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("vanta-research/atom-v1-8b-preview", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("vanta-research/atom-v1-8b-preview")

messages = [
    {"role": "system", "content": "You are Atom, a collaborative thought partner."},
    {"role": "user", "content": "How do neural networks learn?"}
]

inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=512, temperature=0.8)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## Model Details

- **Developed by:** VANTA Research
- **Model type:** Causal language model
- **Base model:** mistralai/Ministral-8B-Instruct-2410
- **Parameters:** 8B
- **License:** CC BY-NC 4.0
- **Training method:** LoRA fine-tuning
- **Format:** Transformers (FP16) + GGUF (Q4_0)

## Capabilities

Optimized for:
- Collaborative problem-solving
- Technical explanations with accessible analogies
- Code generation and debugging
- Exploratory conversations
- Educational dialogue

## Files

- `*.safetensors` - Merged model weights (FP16)
- `atom-ministral-8b-q4_0.gguf` - Quantized model for Ollama/llama.cpp
- `config.json` - Model configuration
- `tokenizer.json` - Tokenizer files

## License

CC BY-NC 4.0 - Non-commercial use only. Contact VANTA Research for commercial licensing.

## Citation

```bibtex
@software{atom_v1_8b_preview,
  title = {Atom v1 8B Preview},
  author = {VANTA Research},
  year = {2025},
  url = {https://huggingface.co/vanta-research/atom-v1-8b-preview}
}
```