Minsore — We make AI fly.

Minsore builds small, efficient models for coding that are easy to run locally — no clusters, no expensive GPUs.

We believe useful intelligence should fit where people need it: on your laptop, in your tools, in your workflow. Small models. Serious intelligence. Start small, scale only when the task demands it.

Our coding family scales with the task:

• Quill (~0.7B) — fast, lightweight coding assistance
• Pepper (~1.5B) — everyday coding, debugging, tool use
• Symphony (~3B) — most capable, agentic workflows

Local-first, developer-first, OpenAI-compatible API.

⚠️ Pepper and Symphony are currently in training. Weights and demos will be published here once ready. A preview of the Quill model is already available.