GSS-creator/adtc-2026-submission-G1-360M
Type de record:
model
Créateur:
GSS
Hôte:
G1-360M V18.2 is a compact, production-ready enterprise AI model built by Gaston Software Solutions LLP (GSS-LLP), Kampala, Uganda — designed to run fully offline on a laptop and reason about real business systems, enterprise workflows, APIs, data validation, and security controls.
# G1-360M V18.2 — ADTC 2026 Submission
**Team:** Gaston Software Solutions LLP (GSS-LLP)
**Track:** Corporate Enterprise
**Model:** G1-360M_V18.2_PRODUCTION_F16
**Hugging Face:** huggingface.co
**Ollama Registry:** ollama.com
---
**G1-360M V18.2** is a compact, production-ready enterprise AI model built by **Gaston Software Solutions LLP (GSS-LLP)**, Kampala, Uganda — designed to run fully offline on a laptop and reason about real business systems, enterprise workflows, APIs, data validation, and security controls.
> **Runs on 8 GB RAM.** G1-360M V18.2 (GGUF F16) is designed for the standard 8 GB laptop profile. When running through **Ollama desktop**, the model loads efficiently and can operate with as little as 4 GB of available RAM, making it practical for offline enterprise environments on any modern laptop or workstation.
---
## ✅ Submission Checklist
- [x] Repository is **public** on GitHub
- [x] `metadata.json` fully filled in — no placeholder values remain
- [x] `metadata.json` contains exactly **2 test prompts** in the `test_prompts` array
- [x] `download_model.sh` downloads the model to `model/`
- [x] Downloaded file is valid **GGUF F16** format
- [x] `model/*.gguf` is in `.gitignore` — weight file not committed
- [x] `REPORT.md` filled in with full technical writeup
- [x] Model runs entirely **offline** — zero external network calls during inference
- [x] African Use Case Bonus claimed (`african_alpha_claim: true`)
---
## Model Journey
How G1 was built — from proprietary enterprise data to production model:
---
## Training Pipeline
The full training pipeline from source data to G1-360M V18.2 Production F16:
---
## How It Works
G1's inference architecture — from user prompt to enterprise response:
---
## Quick Start
### Option A — Ollama (easiest, recommended for team members)
```bash
ollama run gastonsoftwaresolutions234/g1-360m
```
### Opti …