Large models trained on broad data that can be adapted to many downstream tasks rather than being built for one purpose.
They turned machine learning from a bespoke engineering project into something closer to a utility. That shift is what makes almost every other AI category on this page possible.
Models are pre-trained on very large corpora, then refined with supervised and preference-based tuning. Adaptation happens through prompting, retrieval or lightweight fine-tuning.
Capability gains are now coming as much from post-training, tool use and inference-time compute as from raw parameter count. Open-weight models trail the frontier but close the gap for many tasks.
Large GPU clusters, high-bandwidth interconnect and substantial power capacity.
Agent platform for back-office workflows with human approval built in.
Toronto, Canada · Founded 2023
Inference serving stack aimed at cost per token rather than peak throughput.
Austin, United States · Founded 2022
Designed proteins with automated wet-lab validation in the loop.
Boston, United States · Founded 2021
A serving layer that sends each request to the cheapest model meeting a declared quality threshold rather than defaulting to the largest available model.
Source: Huburb demo record · no external source attached
Recent open releases score closer to frontier systems on structured tool-calling evaluations.
Source: Huburb demo record · no external source attached