PhoneLLM Mini Alpha 2

PhoneLLM Mini Alpha 2 is a compact language model from Pipecat, built on NVIDIA Nemotron 3 Nano and optimized for low-latency, multi-turn voice-agent applications. With 30 billion total parameters and a 262K-token context window, it is designed for accurate tool calling and natural customer-support conversations while retaining context across extended interactions. Mini runs with thinking disabled by default, prioritizing responsive answers for real-time voice workflows such as appointment scheduling, information retrieval, and support routing. Its smaller footprint makes it well suited to deployments where conversational speed and serving efficiency are key priorities.