Route by mission need
The architecture can select hosted or self-hosted models without rewriting the whole workflow.
Model routing and evaluation, under Trunnion's control.
Trunnion is model-agnostic today. The Trunnion LLM program explores how model routing, evaluation, provenance, and release review can be controlled inside the same governed architecture. It does not identify a proprietary public model offer.
Conceptual system view. This is not a live product screen.
The architecture can select hosted or self-hosted models without rewriting the whole workflow.
Licensing, provenance, security, and task-specific evaluation belong in the release decision.
Research activity is not represented as a model endpoint, product entitlement, or performance claim.
This sequence explains the intended relationship between the capability, the control plane, human authority, and release evidence.
The mission and data boundary define the model task.
Provider, hosting, licensing, and data rules constrain selection.
A release-specific router assigns work to an approved option.
Task checks and human review govern consequential use.
The available release record identifies the approved routing and review context.
Illustrative workflow. Exact controls, integrations, and evidence are verified for the deployed release.
This program is being developed in relation to Core and the platform family. A link below does not establish that the capability is available in that product.
No proprietary Trunnion model is marketed as available. Training data, benchmarks, model weights, performance, licensing, and launch timing are not stated without release evidence.
Get started
See the governed control plane run against your workflow, in your environment. From cloud to air-gap.