iChancellor’s AI
A frontier research academy

Pioneering science in intelligence.

iChancellor’s AI unites three frontiers — foundational AI research, embodied physical intelligence, and quantum computation — under one open, scientific mission: to build the tools that let humanity understand and improve the world.

Next generation

Coming this autumn

Three new model classes join the Chancellor family on 11.11.26 — built to act, to move, and to perform.

Coming this autumn · 11.11.26
LAAM

Large Action & Activity Model

Beyond single actions: LAAM understands whole activities — routines, workflows, and plans that span hours, not turns. It learns how tasks compose, so it can pick up yours halfway through.

Follow the launch calendar →
Our conviction

The next great scientific instruments won’t sit in observatories — they’ll reason, move, and compute beyond classical limits.

What we build

Three frontiers.
One academy.

All research areas
01 · Research AI

Action models, in the open

We train, dissect, and release fully open Large Action Models — systems that plan and execute, not just predict text — with weights, data, and training code included, so the scientific community can verify every claim and build on every result.

  • Chancellor-1: an open 70B-parameter Large Action Model
  • Full training-data transparency and audit tooling
  • Interpretability research on scientific reasoning
Explore Research AI
02 · Physical AI

Intelligence that touches the world

Our embodied-intelligence lab teaches machines to perceive, plan, and act in unstructured environments — from dexterous manipulation to field robotics for science, agriculture, and disaster response.

  • RoboMan: our humanoid platform for real work
  • World-model learning from real sensor streams
  • Sim-to-real transfer at laboratory scale
Meet RoboMan
03 · Quantum Computing

Computing past the classical horizon

The QuArc program develops error-corrected quantum hardware and the algorithms to match — targeting quantum advantage on the chemistry, materials, and optimization problems that classical machines cannot reach.

  • 128-qubit superconducting testbed, QuArc-128
  • Logical-qubit error correction below threshold
  • Quantum–AI hybrid algorithms for molecular design
Explore Quantum
Open by default

Science you can check

Every model, dataset, and result we publish ships with the artifacts needed to reproduce it. Openness isn’t a release strategy — it’s the method.

0Peer-reviewed papers / yr
0Models launching autumn 2026
0Qubits, QuArc testbed
0+Researchers & engineers