Skip to main content
MODELS

The models we build

One model is in production today. One is in private beta.

Our models

Galvanize 3.1

In production

A multilingual classifier that flags prompt injection in text an agent reads: tool output, web pages, MCP tool descriptions. Galvanize 3.1 is the model behind the API on every plan since 16 November 2026, and its weights are open on Hugging Face.

Temper

Private beta

Finds vulnerable code and proposes the fix. It is for teams that want a security review of their own code. Built from internal research and training. Access is by request while it is in private beta.