Runars vs Cohere

Born in Toronto. Served from the US.

Cohere is Toronto-born and has just signed to merge with Germany's Aleph Alpha to sell transatlantic sovereignty. But the standard API defaults to a US inference region, North and Compass are custom-quoted, and Command A+ scores 13 on the Artificial Analysis Intelligence Index — below the 19 median for open-weights models, where GLM 5.3 scores 45. Runars serves GLM 5.3 and Kimi K3 from Montreal and Calgary, in CAD, with both OpenAI- and Anthropic-compatible APIs.

Credit where due

Cohere is a serious lab — Toronto-born, with Apache 2.0 open-weights releases, real enterprise customers and a research bench that runs from Aya to Embed 5. If you need a full embed/rerank/parse stack or an air-gapped North deployment, they're a legitimate first call. What they aren't — by default — is inference in Canada.

Data residency

Canadian-founded. US-served.

Cohere was born in Toronto — genuinely Canadian. But its standard API's default inference runs in a US region, and Canada data residency is a private-deployment conversation, not something that comes with an API key. Runars serves every request from Montreal and Calgary, on the public API, no deployment project required.

request trace
  1. 1your app🇨🇦 Canada
  2. 2runars.ca/v1🇨🇦 Canada
  3. 3inference🇨🇦 Montreal · Calgary
  4. 4logs + prompt cache🇨🇦 Canada

border crossings: 0

Billing

Priced in dollars. Canadian ones.

Cohere's API models are listed in USD — Command A runs $2.50 in / $10.00 out per million tokens — while North, Compass and Model Vault are custom-quoted or billed as monthly instances ($2,000–$6,500/mo). Runars bills per token in CAD with the right provincial tax. No instance sizing, no sales cycle, no USD card charge.

invoice.pdf
Usage (tokens)CAD
GST/HSTby province
CurrencyCAD
FX surprise$0.00

Compatibility

Keep your code. Change one URL.

Cohere ships an OpenAI compatibility layer alongside its own SDK — chat, tools and structured outputs work by swapping the base URL. Runars goes one further and also speaks the Anthropic Messages API natively, so Claude Code and Anthropic-SDK agents run against it too.

app.py
client = Anthropic(
  base_url="https://runars.ca",
  api_key=os.environ["RUNARS_API_KEY"],
)

# OpenAI SDK works too — same key.

Head to head.

The short version, row by row. Every Cohere claim links to a source at the bottom of the page.

Canadian inference
Runars
Montreal & Calgary. Logs and prompt cache stay in Canada.
Cohere
Toronto-founded, but the standard API defaults to a US region. Canada residency means a private deployment.
Independent benchmark score
Runars
GLM 5.3: 45 · Kimi K3: 44 on AA Intelligence Index v4.3.2.
Cohere
Command A+ scores 13 on the same index — below the 19 median for open-weights models.
Pay-as-you-go token pricing
Runars
Every model, listed, per token.
Cohere
API models are listed (Command A: $2.50/$10.00 USD). North, Compass and Model Vault are custom-quoted or monthly instances.
Billing currency
Runars
CAD with GST/HST/PST by province. No FX line on your invoice.
Cohere
USD.
Anthropic Messages API
Runars
Native /v1/messages. Point the Anthropic SDK or Claude Code at runars.ca.
Cohere
Not offered. Cohere SDK and an OpenAI compatibility layer.
OpenAI-compatible API
Runars
/v1/chat/completions and /v1/responses.
Cohere
Compatibility layer: chat completions, tool calling, structured outputs.
Search stack (embed, rerank, parse)
Runars
Not offered — Runars is text generation and compute.
Cohere
Embed 5, Rerank 4, Parse 5 and Transcribe. A genuine strength.
Enterprise agent platform
Runars
Claude Code and any OpenAI- or Anthropic-SDK agent, by changing one URL.
Cohere
North 2: agent harness, automations, air-gapped deployment. Custom-priced.
Independent and Canadian-owned
Runars
ARS Technologies, based in Canada.
Cohere
Merging with Germany's Aleph Alpha. Deal pending regulatory approvals.

Same models. In CAD.

Per 1M tokens, input / output. Runars prices are live. Cohere prices were listed in USD on 2026-10-16 and are converted at 1.4196 USD→CAD for comparison.

ModelRunars (CAD)Cohere (≈ CAD)
GLM 5.3$2.25 / $7.08$3.55 / $14.20(Command A)
GLM 5.3 Flash$0.24 / $0.80$0.71 / $2.13(Aya Expanse 32B)

Switch in one line.

Point your existing OpenAI or Anthropic SDK at Runars and keep your code. If we're not the right fit next to Cohere, we'll tell you.

base_url = "https://…"
base_url = "https://runars.ca/v1"

Two minutes. No sales call.