The hybrid model
A routing architecture that owns high-volume, sensitive, predictable workloads and rents bursty, experimental, or frontier-reasoning ones. Each request goes to the cheapest capable option, balancing cost, control, and capability.
Route by workload
A gateway layer sends routine, private, high-volume tasks to self-hosted models and hard reasoning or peak load to APIs. Because open-weight quality is within ~3–5% of frontier, most volume can run cheaply on owned infrastructure.
The hybrid model turns the owned-versus-rented debate into an optimization instead of an argument. Baseline capacity runs on owned hardware sized to median load; spikes and the hardest reasoning burst to rented frontier models. Instrumenting spend by transaction and workflow keeps the routing honest, so the split reflects real economics rather than a preference.
Frequently asked questions.
Is hybrid harder to operate than picking one?
It adds a routing layer, but abstraction tools make model backends swappable. The operational cost is usually worth it because hybrid captures owned-scale savings without giving up frontier capability or burst flexibility.