---
tags:
- Coding
- Developmental
- Multimodal
- GGUF
- Q_4_KM
- Foundation
license: apache-2.0
language:
- en
base_model:
- Qwen/Qwen2.5-Coder-1.5B
pipeline_tag: image-text-to-text
---


![Folio-1](folio.png)


![Status](https://img.shields.io/badge/status-multimodal-blue) ![Params](https://img.shields.io/badge/parameters-1.5b-blue) ![Status](https://img.shields.io/badge/company-Foundation-black) 


# Introducing the first coding model (1.5B), in joint development between *Foundation* and *AeroAI*
**Trained on Foundation's *Forge* on MLX**

**Available in Q4_K_M GGUF**


<video controls autoplay muted loop playsinline width="720">
  <source src="https://huggingface.co/AeroAIAviation/Folio-1/resolve/main/video.MP4" type="video/mp4">
</video>



## Intended Use

- Coding and developmental work.

  
## Limitations and Safety

> **Not for commercial and/or operational use.** Foundation's ***Folio-1*** is experimental and may produce incorrect, outdated, or fabricated answers.

- Do **not** use it for commerical or cyber work. 
- Small models may hallucinate, particularly on numeric and regulatory detail.



## Built in development between Foundation and AeroAI.
![foundation](foundationfolio.JPG)


## Quick Start - Ollama 

```
ollama run hf.co/AeroAIAviation/Folio-1:Q4_K_M
```

### Python (Transformers)


```bash
pip install -U transformers torch accelerate
```

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "AeroAI/Folio-1"
tok = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

messages = [{"role": "user", "content": "Code me a single file website. You decide the context and domain."}]
inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
out = model.generate(inputs, max_new_tokens=256, temperature=0.7, do_sample=True)
print(tok.decode(out[0][inputs.shape[-1]:], skip_special_tokens=True))
```

### Apple Silicon (MLX)

```bash
pip install -U mlx-lm
mlx_lm.generate --model AeroAI/Folio-1 --prompt "Explain Terminal shell commands and what they are." --max-tokens 256
```

## Model Details

![Foundation](foliodesign.png)

| Field | Value |
|---|---|
| Developer | Foundation + AeroAI |
| Family | Folio |
| Parameters | 1.5B |
| Type | Causal language model (chat) |
| Domain | Coding and developmental |
| Input / Output | Text and image/ Text |
| Context length | 32,000 |
| Training data | AeroAI x Foundation Proprietary + Distillation |
| Fine-tuning method | LoRa + FFT |
| License | Apache 2.0 |


## Benchmarks and performance (coding)
| Benchmark | Score (%) |
|---|---:|
| HumanEval | 43.9 |
| HumanEval+ | 36.6 |
| MBPP | 69.2 |
| MBPP+ | 58.6 |
| MBPP (3-shot) | 59.2 |
| BigCodeBench-Complete (Full) | 34.6 |
| BigCodeBench-Complete (Hard) | 9.5 |



## Training
Folio-1 was trained on Foundation's ***FORGE*** 
The model was trained in four stages: 
1. Base model testing.
2. Model training.
3. Model testing.
4. Post training/fine tuning and alignment.

Training materials included AeroAI X Foundation's proprietary datasets, along with distilled datasets using Perplexity, OpenAI's ChatGPT, Oscilla, and Anthropic's Claude. 






## Citation

```bibtex
@misc{aeroai_Folio-1,
  title  = {Folio-1: An Experimental Coding and Developmental work model},
  author = {{AeroAI}},
  year   = {2026}
}
```

## Contact/Support
Website: https://huggingface.co/AeroAIAviation/Folio-1 https://foundationai-iota.vercel.app/models
Email: aeroaiaviation@icloud.com or aeroaisupport@icloud.com