PhoneLLM Max Alpha 2
PhoneLLM Max Alpha 2 is Pipecat’s larger open-weight language model for voice agents, with 120B total parameters, 12B active per token, a one-million-token context window, and microthinking for tool use.
Model
PhoneLLM Max Alpha 2 is a full-parameter supervised fine-tune of the language backbone of NVIDIA Nemotron 3.5 Super VL. The vision encoder and projector were removed before training, making Max a text-only model designed for phone conversations and tool calling.
Max introduces microthinking: a single short sentence of reasoning before each response. This gives the model a brief planning step before replying or selecting a tool, without generating the extended reasoning typical of thinking models.
Alpha 2 also expands training across prompting styles, tool sets, and synthetic Pipecat Flows conversations. Max is the larger variant for voice-agent workflows that prioritize accuracy and more demanding conversational decisions. Full details are in Pipecat’s announcement and the model card.
Dedicated Endpoint
DocsA Dedicated Endpoint is your own deployment of PhoneLLM Max 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-max-alpha-2