Release notes

Use dynamic variables to set the voice per call, customizing the assistant voice for each end user without creating separate assistants.

Kimi K2.6 by Moonshot AI is now available on the Telnyx Inference API through the OpenAI-compatible Chat Completions endpoint. No new endpoints or SDK updates required.

Voice AI Assistants now support Deepgram Keyterm Prompting for Nova-3 and Flux, boosting recognition accuracy for up to 100 important terms, product names, and industry jargon.

Run LiveKit on Telnyx to reduce costs, minimize latency, and add enterprise-grade telephony to your voice AI agents.

Define how words are pronounced once and apply it across voices and providers. No re-tuning per voice. No per-provider workarounds.

Inworld's TTS is now available on Telnyx, delivering professional voice actor quality at up to 90% cost savings.

Telnyx Voice AI Assistants now support Anthropic Claude Haiku 4.5 as a native model option. Developers can deploy the model without managing Anthropic API keys, while still retaining the option to...

Telnyx’s importer now handles Retell multi-state agents as native multi-assistant graphs, improving workflow stability and context continuity.

Telnyx has deployed GPUs in Sydney, enabling ultra-low-latency, onshore Voice AI Agents that deliver natural, human-like conversations for Australian businesses.

Telnyx Voice AI Agents now support warm transfers, allowing real-time handoffs between agents with conversation context included to avoid repeated information and improve call flow efficiency.

Telnyx now supports GPT-5 and Groq gpt-oss-120b for Voice AI Agents, delivering more choice, faster responses, and cost-efficient performance for real-time voice.

The Telnyx TypeScript/React Library gives developers full control to build custom Voice AI Agent widgets with low-latency, multimodal support.