Canadian too. Graded by its own teacher.
Augure's October battery is real transparency: 230 Canadian-work tasks, published sample sizes and confidence intervals. But the scenarios are Augure's own, and the models Augure recommends — Rosedale One and Ossington 5 — remain unscored by any independent index. Its new Augure Open preview also serves GLM-5.3, the one model on this page third parties have measured — as a limited preview with no SLA. Runars serves the same published models from our own GPUs in Canada, in general availability, billed in CAD.
Credit where due
Credit twice over: Augure published honest October numbers — sample sizes, confidence intervals, and a candid admission that its flagship trails the Claude reference it picked — and the Augure Open preview is candid about its own terms: no SLA, prices may change, limited time. Weights on their own GPUs in Canada is a real commitment. It doesn't change that every recommended score on their page was authored, run and graded by Augure.
Data residency
Zero border crossings.
Augure's new Augure Open preview tightens its story: open models on Augure's own GPUs, weights hosted by Augure, and a call that fails rather than leaves the country. Runars is the same design — our own GPUs in Montreal and Calgary, weights we serve ourselves, logs and prompt cache on Canadian soil. Two Canadian providers, same architecture. The difference isn't geography; it's everything else on this page.
- 1your app🇨🇦 Canada
- 2runars.ca/v1🇨🇦 Canada
- 3inference🇨🇦 Montreal · Calgary
- 4logs + prompt cache🇨🇦 Canada
border crossings: 0
Performance
Scored in public. Or not at all.
On the two benchmarks Augure published that independent labs have also run, their numbers trail ours: Rosedale One posts 83.0 on GPQA Diamond and 86.0 on MMLU-Pro — self-run — while Vals AI measures GLM 5.3 at 88.1 and 86.8, and Kimi K3 at 92.9 and 88.0, on the same named benchmarks (different harnesses; read the gaps as directional). And the instrument matters: GLM 5.3 scores 45 on the Artificial Analysis Intelligence Index, a third party with a published method. Augure's open preview now serves that same published GLM-5.3 — the scores belong to the model, not to anything Augure built — while its own materials point customers to Rosedale One and Ossington 5 for capability: the two models no independent index has scored.
GPQA Diamond
- GLM 5.3
- 88.1
- Rosedale One
- 83.0
MMLU-Pro
- GLM 5.3
- 86.8
- Rosedale One
- 86.0
AA Intelligence Index
- GLM 5.3
- 45
- Augure models
- no entry
runars scores: vals ai + artificial analysis (independent, published method) · augure scores: october battery (self-run, self-graded)
Compatibility
Keep your code. Change one URL.
Augure's API is OpenAI-compatible — chat completions, embeddings, and their own Augure Code CLI. Runars also speaks the Anthropic Messages API, so Claude Code and Anthropic-SDK agents work by changing one base URL.
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 Augure claim links to a source at the bottom of the page.
Same models. In CAD.
Per 1M tokens, input / output. Runars prices are live. Augure prices were listed in CAD on 2026-10-22 — no conversion.
The first row is the same published GLM-5.3 we serve — their preview price against our live GA price. The second pairs their open preview's Qwen flash against the flash tier we already run. Rows three and four are Augure's own models, priced by Augure, scored by Augure.
| Model | Runars (CAD) | Augure (CAD) |
|---|---|---|
| GLM 5.3 | $2.21 / $6.96 | $3.00 / $7.00(GLM-5.3 — Augure Open (open-glm-5-3)) |
| GLM 5.3 Flash | $0.17 / $0.58 | $2.00 / $4.00(Qwen3.8-Flash-Next — Augure Open (open-qwen-3.8-flash-next)) |
| GLM 5.3 | $2.21 / $6.96 | $2.50 / $6.00(Rosedale One) |
| GLM 5.3 Flash | $0.17 / $0.58 | $1.50 / $3.00(Ossington 5) |
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 Augure, we'll tell you.
base_url = "https://…"
base_url = "https://runars.ca/v1"Two minutes. No sales call.
Sources · checked 2026-10-22
- Augure — model reference (Augure Open preview)
- Augure — benchmarks (October 2026)
- Augure — benchmark results JSON
- Augure — homepage
- Augure — developers
- Augure — Rosedale One
- Augure — Ossington 5
- Artificial Analysis — GLM-5.3
- Artificial Analysis — Kimi K3
- Artificial Analysis — Intelligence Index methodology
- Vals AI — GLM 5.3 (independent runs)
- Vals AI — Kimi K3 (independent runs)
Augure is a trademark of its owner and is referenced here for comparison only. Competitor details change; check Augure's site for current terms. Spotted something out of date? Tell us and we'll fix it.