Providers
The providers we work across, and our takes on the models
Already committed to a provider? We build on yours. Don't want to manage one? We host and run it for you. Below: the providers we work across, the models we track, and our published takes on each — added to as the model layer moves.
Providers, models, takes
A take is our published judgment on a model or platform — what it changes, what to test before you rely on it, where it doesn’t fit. Takes are dated: models move, and a take is a snapshot. Where a claim is checkable, the take carries its source.
Anthropic
Maker of the Claude model family. Our primary internal stack today — most of our production builds run on it, which is why our takes here are the most detailed.
Claude Opus 4.7
The most capable tier — long-horizon agentic work, complex analysis, sustained multi-step workflows.
Our take →Claude Sonnet 4.6
The production workhorse — the capability-to-speed balance most agent workloads run on.
Claude Haiku 4.5
Fast and cost-efficient — high-volume classification, extraction, and real-time interaction.
Sierra
Agent-platform company, not a model maker — included because the buy-a-platform path is the question many teams are actually weighing. Our take covers what's under the hood and when it fits.
Where it runs
Two paths, both first-class. The model layer is a choice you keep — not one we make for you.
Bring your provider
Your keys, your contracts, your data residency. We build on the provider your governance already approves.
Or we host it for you
We stand up and run the model layer — deployed in Australian regions on AWS, Google Cloud, or Microsoft Azure, with audit trails for compliance.
Compliance roadmap — Diligence in dates
Diligence is the fourth D and the first thing procurement asks about. Here’s where we are, and where we’re headed — without promising a posture we don’t hold yet.
Weighing the model layer for a real workflow?
Bring the workflow — we'll talk through what fits. 30-min call books straight to a real engineer.