Alchymia AI Develops Foundational Models for High-Level Intelligence at Lower Compute and Energy Costs

Alchymia AI’s Regent model is an open-source alternative to transformer architecture, maintaining a fixed memory footprint and selectively retrieving information only when required.

Regent is lightweight enough to run locally on consumer hardware, eliminating the need for hyperscale cloud clusters.

The company’s latest frontier model, Darkhorse, uses adaptive computation, dynamically scaling its processing depth to match the complexity of a prompt, lowering overall inference costs.