CanXP AI has launched MaplePT, a small language model (SLM) trained entirely in Canada, marking what the company calls a step toward “sustainable, sovereign AI” that is accessible to all Canadians.

Built and hosted exclusively within Canada, MaplePT was developed using CanXP’s distributed, low-energy training framework. The company says the system matches competitive reasoning and alignment performance while running on consumer-grade graphics processing unit – an approach that reduces the need for billion-dollar data centres or foreign cloud infrastructure.

“Canada doesn’t need gigawatt-scale AI data centres that strain our grid and environment just to satisfy the compute demands of foreign markets,” said Vince McMullin, co-founder and CEO of CanXP AI. “What we need is capital investment into applied computing, usable systems, and efficient models that every Canadian—from academics to professionals—can use.”

Based in Saint John and Toronto, CanXP launched in September. The company was co-founded by McMullin and Eric Hubacheck, who leads partnerships and commercialization efforts. The 15-member team includes contributors and advisors across infrastructure, AI training, and product development, with legal support from Toronto law firm Osler, Hoskin & Harcourt LLP.

MaplePT is the company’s flagship model and part of a broader effort to build a Canadian federated AI network that connects models across research, education, and industry. Through the CanXP platform, organizations can deploy their own SLMs, keep them private or public, and contribute to a network governed under Canadian law and powered by clean energy.

McMullin said CanXP’s approach is a response to Canada’s reliance on large foreign tech firms. The company’s technology, he said, demonstrates that “we can support Canadians and Canadian businesses without involving U.S. big tech” or building “diesel-guzzling AI data centres in our backyard.”

CanXP’s roots trace back to early AI development work for General Motors before political and commercial shifts in the U.S. curtailed cross-border partnerships. “When U.S. companies pulled back, we decided to dig in and build for Canadians,” McMullin said. “Our goal is to make AI accessible to everyone—from students to professionals.”

The company said it has already attracted paying customers, including professionals and early enterprise adopters using its AI platform for research, writing, and privacy-sensitive applications. All data, McMullin emphasized, remains hosted within Canada to ensure compliance and residency.

CanXP operates on a subscription model, with tiers designed for students, professionals, and institutions. Enterprise licensing is also available for organizations requiring locally hosted AI solutions.

The company’s first publicly released model, MaplePT-Mini (v1), is available under an MIT license on Hugging Face and Docker Hub, accompanied by an open-access research paper. CanXP is now seeking partners across healthcare, education, legal, defence, and manufacturing sectors as it expands its network.

Current Investors include the New Brunswick Innovation Foundation, Business Development Bank of Canada, Golden Triangle Angel Network in Waterloo, and several smaller investors nationwide. CanXP is currently raising additional funding to extend its federated AI network.

“MaplePT proves that sustainability and sovereignty are the same fight,” said McMullin. “Canada can lead with models trained responsibly at home, governed by our laws, and accessible to every citizen and small business that wants to participate.”