Enterprise
At 10,000 AI calls a day, your AI gets its own GPUs.
High-volume teams run their AI on dedicated GPU nodes. Each node handles about 2,000 to 3,000 calls an hour, and the GPUs are part of your platform fee, so there is no charge per minute.
- 10,000+
- AI calls a day: where dedicated GPUs start
- 2,000–3,000
- calls a GPU node handles in an hour
- 0
- charges per minute or per call
- 1
- flat platform fee, GPUs included
Sizing
How many GPUs your volume needs.
GPU nodes run through your calling hours, so they are sized from your busiest hour, not your daily total.
Sized for your busiest hour, taken as 30% above the average. A node handles about 2,000 to 3,000 calls an hour; this uses 2,500.
You’d run
2 GPU nodes
Busiest hour
2,889 calls
At this volume your nodes run at about 58% in the busiest hour. They’re part of your platform fee, with no charge per call or per minute.
Your runtimes, in your workspace
Dedicated GPU pipelines appear next to any Gemini Live runtime you use. Pick which one each dial session and inbound number uses, and see each one's status and recent calls.
- A session pauses itself if its runtime goes offline, and tells you why.
- Inbound numbers with an offline AI are flagged on the routing page.

Sessions sized for the volume
Run sessions at 100 lines, rotate local caller IDs and pace calls to each carrier's limit. Retries, skip rules and auto-pausing keep the volume clean.

What's included
- Dedicated GPU nodes
- Speech recognition, the language model and the voice run on hardware that handles only your calls.
- No per-minute charge
- The GPUs are part of your platform fee, sized on your volume.
- Your carrier
- Bring your own SIP trunk and numbers. Your carrier bills you directly, at its rates.
- The whole platform
- Dialer, AI agents, dial sessions, AI Messenger, the SMS inbox, inbound routing, call log and analytics.
- Several runtimes
- Run more than one AI runtime and choose which one each session or number uses.
- Runtime health
- Each runtime shows when it's online, and sessions pause themselves if it goes offline.
Planning 10,000 calls a day or more?
Tell us your daily volume and calling hours. We'll size the GPU nodes and quote the platform.