Cambridge, Massachusetts robotics startup Tutor Intelligence has rolled out a second generation of its warehouse-focused robots and a proprietary foundation model designed to run them both — Cassie, a mobile single-arm workhorse able to lift up to 50 pounds and move thousands of pounds of goods across a warehouse floor; Sonny, a dual-armed semi-humanoid tuned for shelf work, bin picking and object manipulation; and Ti0, a 4.5-billion-parameter vision-language-action (VLA) model trained inside a 100-robot classroom the company calls Data Factory 1.
Data Factory 1 turns robots into their own teachers
The Data Factory 1 facility runs a fleet of Tutor robots 24/7 through fresh manipulation tasks, generating in-the-wild trajectories that Tutor uses to fine-tune Ti0. Rather than lean on internet-scale text and video, Ti0 is trained primarily on this first-person robot data — a bet that a "physical" foundation model needs data collected under load, not scraped from the web.
"Generally capable and immediately useful"
"We're in this unique position of building robots that are both generally capable and immediately useful," said co-founder and CEO Josh Gruenstein. Tutor, founded by MIT alumni in 2023, closed a $34 million Series A in December 2025 and has been deploying pilot Cassie units into contract manufacturing and third-party logistics customers over 2026.
Ti0 joins a fast-growing club of robot foundation models
Ti0 lands the same week that Reward AI unveiled OM-1, a human-data-only robot foundation model, and just after Noetive raised $41M for its own industrial "world model" for factories and logistics. Tutor's bet is closer to the humanoid pack: single-model, multi-embodiment control targeting real warehouse throughput, in a market where Amazon's Fauna deal and Agility, Figure and Apptronik commercial rollouts have set a rapid pace.
Reporting based on coverage from SiliconANGLE and Tutor Intelligence corporate disclosures.