Overview
Gemini Robotics ER 2 Preview is a preview vision-language model for embodied reasoning in robotics, built on Gemini 3.5 Flash. It interprets multimodal information about physical environments and can help plan tasks and coordinate robots and tools. (docs.cloud.google.com)
Release date: July 30, 2026, when Google announced the endpoint in public preview. This is the Gemini API release date; Google Cloud separately lists October 27, 2026 for its platform release. (aistudio.google.com)
Capabilities and uses
- Accepts text, images, video and audio; outputs text. The documented input limit is 131,072 tokens and the output limit is 65,536 tokens. (ai.google.dev)
- Supports spatial reasoning, pointing, instrument reading, video moment finding and task-progress classification. It can orchestrate multi-step tool use and coordinate multiple robots, including through external robot functions or vision-language-action models. (aistudio.google.com)
- Gemini API capabilities include function calling, code execution, structured output, thinking, caching and Batch API. This standard endpoint does not support the Live API, image generation or audio generation; the separately named streaming preview is a different endpoint. (ai.google.dev)
Limitations and safety
This is a preview model, not a guarantee of correct perception or safe physical action. Google says users are responsible for maintaining a safe robot environment and should not use the robotics models for safety-critical applications such as healthcare or transportation where malfunction could cause death, injury or property damage. (docs.cloud.google.com)
Gemini Developer API pricing
Official USD prices, checked November 16, 2026. The following applies to this standard endpoint on the Gemini Developer API; other platforms may have different terms and prices. Token prices are per 1 million tokens. Output prices include thinking tokens. (aistudio.google.com)
| Tier | Through December 31, 2026 | From January 1, 2027 |
|---|---|---|
| Standard paid input, including text, image, video and audio | $1.00 | $2.00 |
| Standard paid output | $5.00 | $10.00 |
| Batch paid input, including text, image, video and audio | $0.50 | $1.00 |
| Batch paid output | $2.50 | $5.00 |
| Standard paid cached tokens | $0.10 | $0.20 |
| Batch paid cached tokens | $0.05 | $0.10 |
| Cache storage, per 1 million tokens per hour | $0.50 | $1.00 |
The Standard free tier lists input and output as free, but caching as unavailable. Batch is unavailable on the free tier. Google also lists separate Search grounding charges: 5,000 free search requests per month shared across Gemini 3 and newer models, then $14 per 1,000 requests, subject to the stated daily cap. (aistudio.google.com)