PhoneLLM Mini Alpha 2
PhoneLLM Mini Alpha 2 is Pipecat’s open-weight language model for low-latency voice agents, with 30B total parameters, 3.5B active per token, and a 256K-token context window.
Model
PhoneLLM Mini Alpha 2 is a full-parameter supervised fine-tune of NVIDIA Nemotron 3 Nano, designed for phone conversations and customer-service agents. Its hybrid Mamba-Transformer Mixture-of-Experts architecture prioritizes fast responses, with thinking disabled to avoid a reasoning delay before each reply or tool call.
Alpha 2 expands training across different use cases, prompting styles, and tool sets. It also includes synthetic Pipecat Flows conversations, helping the model follow structured workflows across multiple turns.
Mini is the smaller PhoneLLM Alpha 2 variant, suited to voice-agent applications where response speed and model size are the main priorities. Full details are in Pipecat’s announcement and the model card.
Dedicated Endpoint
DocsA Dedicated Endpoint is your own deployment of PhoneLLM Mini Alpha 2, on an inference stack that Modal tunes and autoscales. Bring fine-tuned weights if you have them. Compute is billed by the second, and only while it runs.
modal endpoint create --model pipecat-ai/phonellm-mini-alpha-2