Tesla 1 is coming: our state-of-the-art model for enterprise security and agentic work

Status update, September 9, 2026: Tesla 1 remains in development. It has not been released and there is no announced release date. The original preview follows.

Why we built it

The enterprises we work with need models they can trust with their data, ones that run inside their own walls and never send a token to someone else's cloud. Open-source models could do this, but only if you stood up racks of expensive GPUs to host something massive.

So we built the opposite. Tesla 1 is small enough to run on modest infrastructure, yet genuinely smart and tuned for real agentic work. It is a state-of-the-art model that puts secure, on-premise AI within reach without the GPU bill that usually comes with it.

What we can share so far

• Small and efficient, light enough to run on your own infrastructure without a large GPU cluster.
• A large context window, so entire codebases, document sets, and long agent histories fit in a single pass.
• Trained for far less money than any other frontier model while keeping the same level of performance.

This is just a teaser

Benchmarks and a release date have not been announced. You can register your interest, but access is not currently available.