Tessera Start chatting

Tessera is frontier-scale AI running
on one homebuilt PC.

A 744-billion-parameter model, served from a secondhand PC. No cloud, no API keys. It thinks slowly, and it's worth the wait.

744B
Parameters in
the model
0B
API keys, servers
or subscriptions
0
Of every answer computed
on one homebuilt PC
0%

[ How it fits ]

Most of the brain
stays asleep.

your question
NVMe read activityidle
›
each square = one expert (illustration)0 / 128 awake

Questions that deserve
a big brain.

TRY ASKING “Walk me through how a Mixture-of-Experts router picks which experts fire.” Ask this → WATCH IT THINK See a question travel through the experts, start to finish. TRY ASKING “Review my ESP32 firmware and tell me what will break first.” Ask this → TRY ASKING “Plan a weekend build I can finish with parts I already own.” Ask this →

Ask something that
deserves a slow answer.

One question at a time. Every word is computed in the room, on hardware you could buy secondhand.

Start chatting