Google adds Gemini Robotics ER 2 models to the Gemini API

Google expanded the Gemini API’s embodied-reasoning (robotics) model family on 2026-07-30, per the page’s own “Last updated” timestamp. Gemini Robotics ER 2 Preview is a vision-language model endpoint that helps robots understand their environments precisely, adding agentic orchestration of robots, video progress understanding, multi-robot collaboration, and advanced spatial reasoning on top of what ER 1.6 offered. A Gemini Robotics ER 2 Streaming Preview variant optimizes the same capability for real-time text streaming via the Live API, supporting bidirectional streaming with function calling.

Both new models are priced at $2.00 per 1M input tokens (text, image, video, audio) and $10.00 per 1M output tokens — exactly double the original Gemini Robotics ER 1.6 Preview rate of $1.00/1M input (text/image/video), $2.00/1M input (audio), and $5.00/1M output. ER 1.6 Preview remains on the price list unchanged, so buyers now choose between a cheaper baseline model and a pricier, more capable ER 2 generation rather than a straight replacement.

The same pricing-page update cycle also published four new Gemma 4 fine-tuning SKUs on Vertex AI (E2B IT at $1.40, E4B IT at $1.73, 26B A4B IT at $2.58, and 31B IT at $9.06 per 1M training tokens), and reorganized the Context Cache Storage reference table into grouped rows by model family without changing any underlying per-model rate. Headline Gemini API and Vertex AI token prices for the general-purpose Flash/Pro model ladder, and the consumer Google AI Plus/Pro/Ultra subscription lineup, were verified unchanged in the same capture cycle.