| license: apache-2.0 | |
| tags: | |
| - hackathon | |
| - local-ai | |
| - hp-zgx-nano | |
| # HP-Morningstar-Hackathon | |
| Model collection staged for the HP + Morningstar AI Hackathon, designed to run locally on the HP ZGX Nano AI Station. | |
| ## Models | |
| | Model | Size | Use Case | | |
| |-------|------|----------| | |
| | Qwen/Qwen3-14B-AWQ | 14B params (4-bit) | Agent LLM with strong structured output | | |
| | Qwen/Qwen3.6-27B-FP8 | 27B params (FP8) | Higher-capability agent for complex reasoning | | |
| | mistralai/Ministral-3-8B-Instruct-2512 | 3.8B params | Lightweight/fast agent for rapid iteration | | |
| ## Purpose | |
| These models are pre-staged for a 4-hour hackathon where Morningstar developers build a multi-tool financial compliance agent running entirely on local infrastructure. The models were selected for compatibility with the ZGX Nano's 128GB unified memory and Ollama's serving capabilities. | |
| ## Usage | |
| Clone this repo to the ZGX Nano and serve via Ollama or load directly with transformers/vLLM. | |