README.md

metadata

title: README
emoji: 🔥
colorFrom: indigo
colorTo: gray
sdk: static
pinned: false

Vertex AGI

Small, focused fine-tunes and distillations, built end-to-end — data generation, training, evaluation, and publishing — on consumer hardware.

Model families

Family	Base	Focus
Amethyst	Gemma 3 4B / Llama 3.1 8B	General-purpose chat & instruction-following
Copal	Gemma 3 4B IT	Agentic, tool-calling
Prism Creative	Qwen3-4B	Creative writing, distilled from GLM 5.2
Prism Roleplay	Qwen3-8B	Character roleplay, distilled from GLM 5.2 / Nemotron-3-Nano

Every model repo ships both MLX and GGUF formats in one place, so it runs on Apple Silicon (mlx-lm) or anywhere llama.cpp and its ecosystem (Ollama, LM Studio, etc.) reach.

Links

• Models
• Datasets