The standard configuration runs inference inside your environment. Document content stays where it is deployed—no prompt or document content leaves your infrastructure unless you turn on the premium option below.
Self-hosted by default.
Not a claim about scale.
Company Brain does not market a large model catalogue, because that is not where its value sits. Inference runs self-hosted by default; a premium hosted model is optional, off unless you enable it, and every governance guarantee elsewhere on this site holds either way.
TWO REAL OPTIONSSelf-hosted.
Self-hosted.
Or opt in.
An optional hosted-model path is available. When enabled, prompts are sent to that provider—this is a real trade-off, stated plainly, not hidden in a settings page.
Grounding verification, evidence citation and the review gate on published knowledge apply regardless of which inference path is active. Model choice does not change what is allowed to reach a published state.
We are not competing
on model count.
If what you need is broad model selection for general-purpose tasks, that is not this product's focus. Company Brain's differentiation is the knowledge-governance layer around whichever model answers—see the full deployment scope for the exact data-flow discussion.
See security & deployment →