Open-source inference on NVIDIA B300 GPUs we own

Run a curated lineup of open-source models on B300s Telnyx owns and operates around the globe, so no cloud margin gets stacked into every token.

What's live

↳ Owned hardware: B300 GPUs Telnyx operates, not rented capacity.

↳ Zero data retention: prompts and responses aren't stored after the response returns.

↳ OpenAI-compatible: swap your base URL and keep your existing code.

Learn more ↗︎

AI is leaving the screen

Intelligence is moving off the screen and into vehicles, drones, robots and industrial equipment. Grit Daily looks at what that shift demands from the infrastructure underneath it. Identity, connectivity, compute and communications have to work as one system, close to where the action happens.

Learn more ↗︎

AI Gateway for every team in Mission Control

Give each team its own gateway with approved models, a spend cap over 1, 7 or 30 days, and guardrails that flag or block secrets and card numbers in prompts and responses. Every person gets a scoped inference key, so nobody shares a master API key.

Learn more ↗︎

Decision Models (Beta) on Telnyx Inference

Route tickets, flag incidents and score urgency with typed answers instead of parsing free text. One TypeSafe-compatible request evaluates shared context against up to 64 named questions, mixing choice, yes/no and score types.

Learn more ↗︎

OpenAI GPT-Live in Voice AI Assistants

OpenAI's full-duplex speech-to-speech model listens and speaks at the same time, so callers can interrupt mid-sentence. Telnyx supplies the number, carries the call on its own carrier network and runs the tool layer, with tool calls handled through delegation.

Learn more ↗︎

Rotate DKIM keys through the Email API

Rotate DKIM keys for custom sending domains without opening a support ticket. The API returns the replacement DNS record to publish, so your team controls the full rotation cycle.

Learn more ↗︎

We shipped more than we could fit in an email.