Frontier-class models behind one OpenAI-compatible API. GLM-5.3 and DeepSeek-V4-Pro both run on our Canadian fleet today, with HARI routing every request to the right one, alongside a growing catalogue of open models - Qwen3, Qwen2.5-Coder, QwQ, gpt-oss, Mistral, Phi and more - that load on-demand, plus vision, embedding, image, video, speech, and reranking. Point your existing SDK at ai.spuric.com and switch, your prompts never cross the border.
Every model below is metered by the token and served from inside Canadian jurisdiction. No US Cloud Act exposure, no cross-border data transfer.
Frontier-class reasoning and chat with a massive context window, ideal for long documents, agents, and tool use. Frontier quality without paying frontier-cloud markup.
DeepSeek's frontier Mixture-of-Experts model for deep reasoning, agentic workflows, and tool use, served sovereignly in Canada alongside GLM-5.3. Two frontier flagships, one OpenAI-compatible endpoint. Learn more →
Llama 3.3 70B, DeepSeek-V4-Flash, and a DeepSeek-R1 reasoner for everyday assistants, summarisation, analysis, and Q&A, tuned for low latency on our fleet.
Qwen3-Coder 30B for code generation, completion, and review - for building copilots and developer tooling on sovereign infrastructure.
High-quality text embeddings for semantic search, retrieval-augmented generation, and clustering, the backbone of document Q&A.
Generate and edit images from text with FLUX and SDXL diffusion models, for marketing, product, and design workflows.
Turn prompts into short video clips with Wan 2.1 text-to-video, submitted as asynchronous jobs and delivered to your account.
Send an image with your prompt and the model sees and reasons about it - describe, extract, and analyse charts, screenshots, and documents. OpenAI-compatible vision messages.
Text-to-speech that speaks responses aloud, and speech-to-text that transcribes audio with high accuracy. Drop-in OpenAI-compatible audio endpoints.
Reorder search results and RAG candidates by true relevance with a cross-encoder reranker - sharper retrieval for document Q&A and agents.
You should not have to pick a model. HARI, SPUR's assistant, reads each request and routes it to the best model or pipeline automatically, chat, code, image, or video, so you get the best result without managing a model zoo. Try it free in the browser, then call the same engine over the API.
Need something private? We fine-tune an open model on your data and host it on dedicated capacity, or run your own weights and adapters for you. The model stays in your tenancy, the data stays in Canada, and you reach it through the same API as everything else.
Get an API key and call frontier models hosted in Canada, or talk to us about a private deployment. No quote-walls, no border.
