ServiceNow AI Releases Apriel Voice 27B Preview Multimodal Model
ServiceNow AI has released Apriel Voice 27B Preview, an open multimodal model built on a Gemma backbone that supports speech understanding, complex reasoning, image inputs, and tool calls. This release provides the open-source AI community with a competitive two-stage voice model that pairs a speech encoder with a powerful reasoning backbone, narrowing the gap with proprietary real-time voice systems. The model connects a Whisper large-v3-turbo speech encoder to a Gemma 4 26B-A4B backbone via a learned projector, achieving a 1.91-second mean response latency and strong benchmark performance.
## BACKGROUND
ServiceNow AI builds advanced open-weight foundation models designed for enterprise tasks, reasoning, and multimodal interactions. Gemma is a family of lightweight, state-of-the-art open models developed by Google, sharing technical lineage with their larger Gemini models.