// AI Platform Drift

PhoneLLM Mini voice model now available on Cloudflare Workers AI

Cloudflare developer changelog · published 2026-10-16 11:27 UTC
WORTH KNOWING VERIFY_ONLY MODELAPI
WHAT CHANGEDCloudflare added @cf/daily-co/phonellm-mini-alpha-2 to Workers AI on 15 October 2026. The model is a post-trained NVIDIA Nemotron 3 Nano 30B optimized for voice agents and realtime AI applications. It is accessible through the Workers AI binding (env.AI.run()) or the REST API at /ai/run. Pipecat recommends temperature: 0 and chat_template_kwargs: { enable_thinking: false }.
WHY IT MATTERSVoice agents typically choose between fast models that miss tool calls and slower reasoning-mode models that handle them correctly. PhoneLLM is trained to follow instructions and call tools accurately without reasoning mode, reducing latency in speech-to-text → LLM → text-to-speech pipelines where every second adds to caller wait time. Python Workers are now generally available, making it possible to run Pipecat voice agents on Cloudflare infrastructure.
WHAT TO DOIf you are building voice agents on Workers AI, test @cf/daily-co/phonellm-mini-alpha-2 against your current model to measure latency and tool-calling accuracy. Set temperature: 0 and chat_template_kwargs: { enable_thinking: false } as recommended. Review the PhoneLLM model page and Workers AI pricing to understand cost implications.
Model-reported confidence 85% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## Workers AI, Realtime - Pipecat PhoneLLM Mini now available on Workers AI
+https://developers.cloudflare.com/changelog/post/2026-10-15-phonellm-mini-workers-ai/
+Thu, 15 Oct 2026 00:00:00 GMT
+
+@cf/daily-co/phonellm-mini-alpha-2, a model from the Pipecat ↗︎ team at Daily, is now available on Workers AI. PhoneLLM is built for voice agents and other realtime AI applications, where a model has to reply quickly and call the right tool mid-conversation.
+In a voice agent, the language model sits between speech-to-text and text-to-speech, so every second it takes adds to the caller's wait. Developers usually choose between a fast model without reasoning mode, which can miss instructions or tool calls, and a slower model in reasoning mode, which handles both well. Pipecat post-trained NVIDIA Nemotron 3 Nano 30B on phone and customer support scenarios so PhoneLLM follows instructions and calls tools better without reasoning mode.
+You can use PhoneLLM through the Workers AI binding (env.AI.run()) or the REST API at /ai/run. Pipecat recommends temperature: 0. Also set chat_template_kwargs: { enable_thinking: false }.
+Run a voice agent with Pipecat on Workers
+Pipecat ↗︎ is an open-source Python framework for building voice agents. Now that Python Workers are generally available, you can run a Pipecat voice agent on Cloudflare.
+Our Pipecat voice agent example ↗︎ runs in a Durable Object and uses WebRTC as the voice transport through the Realtime SFU. It uses DeepSeek V4 Flash by default. To deploy it with PhoneLLM, give this prompt to your coding agent:
+Copy promptPrompt copied!
+For more information, refer to the PhoneLLM Mini model page, Pipecat's model card ↗︎, and Workers AI pricing.
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.92, needed 0.5
triage reasoningNew model availability on Workers AI platform with specific deployment guidance and configuration requirements.
signalsnew model added to Workers AI platform · model identifier: @cf/daily-co/phonellm-mini-alpha-2 · specific use case: voice agents and realtime AI · deployment method documented: Workers AI binding and REST API · configuration requirements specified: temperature and chat_template_kwargs · integration with Python Workers and Pipecat framework · availability date: 15 Oct 2026
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)