Why Runars exists: AI compute for Canadians, in Canada

Most teams building with AI in Canada are using the same handful of open models as everyone else. The models are good. The SDKs are familiar. The tooling works. What's usually unclear is everything underneath: where the request physically runs, where the logs are kept, which currency the invoice arrives in, and how much of that you have to ask a sales team about before you can sign anything.

That gap is why we built Runars.

Where the request runs is a real question

Several strong platforms serve the models Canadian developers want. Most of them are very good at what they do. The difference is where the request goes.

• Routers forward your prompt to whichever upstream provider is available. OpenRouter's in-region routing covers the US and EU on its Business and Enterprise plans. Canada isn't among the options it publishes.
• Hosted platforms run their serverless inference in US or EU regions. Canadian residency is available through an enterprise or custom-deployment conversation.
• Some providers don't publish where their data centres are at all.

None of this makes those companies bad choices. It does mean that "where does our data go?" often turns into a procurement project. For Runars, the answer is the default: inference, logs and prompt cache stay in Canada, from Montreal and Calgary, on every plan, starting at zero dollars.

Why Canada specifically

Canadian organisations answer to federal and provincial privacy law, including PIPEDA and, in Quebec, Law 25. Public-sector buyers, health and financial services, and any team handling customer records will eventually ask where inference physically happens and who can access the systems it runs on.

We think those are reasonable questions, and they shouldn't require a bespoke contract to answer. Location is necessary but not the whole picture. Who operates the infrastructure, and which legal regimes could reach it, matter too. We'd rather explain that full picture than lean on the word "Canadian" alone.

Why the invoice matters too

Billing in USD looks like a small detail until finance has to reconcile a US card charge against a moving exchange rate every month. Runars bills usage in CAD, with GST, HST or PST applied for the province on the invoice. There's no FX line to explain.

Card-funded credit top-ups also add fees on some platforms. OpenRouter charges 5.5% on card credit purchases, with a $0.80 minimum. Runars is pay-per-token, with no base fee and no minimum, and usage is invoiced.

Why the API matters

Switching providers shouldn't mean rewriting your application. Runars speaks the OpenAI-compatible /v1/chat/completions and /v1/responses endpoints, and the native Anthropic Messages API at /v1/messages. Point your SDK, your agents or Claude Code at the new base URL with a new key, and the code you already wrote keeps working.

That's the part we care about most. If Canadian infrastructure means changing your stack, most teams won't make the move. If it means changing one URL, they can.

What we're not claiming

Speed is not our headline. Other providers post faster throughput numbers, and some of them are excellent. What we'd point to instead is independent measurement. Our served models are listed on Artificial Analysis, an independent index with a published methodology. On its Intelligence Index v4.3.2, GLM 5.3 scores 45 and Kimi K3 scores 44.

We'd rather you check those numbers than take our word for them.

Try it

If you're evaluating inference for a Canadian team, start with the quickstart or read how we compare on data residency, billing and API compatibility. If we're not the right fit for your workload, the comparison pages say so too.

Questions about residency or the data-flow diagram? Email us at support@runars.ca.