Montréal company, born inside Mila, selling LLM-serving containers that run alongside vLLM in a customer's own cloud and tune GPU operating parameters to cut the energy used per token.
Canadian innovation builds a brighter tomorrow.
Montréal company, born inside Mila, selling LLM-serving containers that run alongside vLLM in a customer's own cloud and tune GPU operating parameters to cut the energy used per token.
Its terms name CarbonForge Systems Inc., a Québec corporation, with an operating address on rue Saint-Urbain in Montréal; REQ 1181725186 is Active. carbonforge.ai
Company benchmark (June 2026): about 30% lower GPU power at equal or slightly higher throughput on saturated decode with vLLM on 2×A100. carbonforge.ai
Names Heavybit, Active Impact Investments and Cycle Momentum as investors; no round or amount is disclosed. carbonforge.ai
Products CarbonForge publishes on its own site. Each card links to the page it was taken from.
LLM inference optimisation
Company describes serving images built for a specific combination of hardware, model and workload, with a scheduler and a controller that adapts GPU operating parameters, running alongside vLLM in the customer's cloud and licensed by GPU class.
Learn moreCarbonForge reported benchmark results showing about 30% lower GPU power at equal or slightly higher throughput on saturated decode, and 13–14% better tokens per watt at matched capacity.
Sourced from the company's own pages or announcements.
Jean-Maxime Larouche
Co-founder and Chief Executive Officer
Pierre-Luc Bacon
Co-founder and Chief Scientist
Laurent Maisonnave
Co-founder and Chief Operating Officer
Terms the company uses about its own work.
Other companies in AI.
Discover companies, track funding, explore career opportunities and stay informed about Canada’s growing tech ecosystem.
