Welcome to Veyra AI

Veyra AI builds tiny English language models for fast local inference. We focus is on compact, CPU-friendly models that are easy to run and fine-tunes.

Our work centers on general base models, function calling, Python-focused variants, distillation, RLVR, tool use, and benchmarks.

Current Model Families

• Veyra2 50M / 30M / 15M — Newer models with a smaller footprint, designed for efficient local inference. These models may not perform as strongly as the original Veyra 30M in all tasks.
• Veyra 30M — A proven 30M base model with strong instruction-following ability and balanced general-purpose capabilities.

Planned Model Families

• Veyra3 50M — Next-generation models optimized for low-latency inference and on-device deployment, with a focus on speed, efficiency, and responsiveness.

Note

Kairo models are experimental and should not be used as reliable general-purpose language models.