Unknown · v2 · since 30. Juli 2026 · 3× · last seen Aug 01, 2026
58
Momentum
Gemini Robotics ER 2 is Google DeepMind's current embodied reasoning model, acting as a high-level reasoning layer for robots. Built on Gemini 3.5 Flash, it processes visual, video, audio and text input, plans multi-step tasks, orchestrates vision-language-action (VLA) models and external tools, and enables multi-robot collaboration. The model was unveiled on July 30, 2026, and is available as "gemini-robotics-er-2-preview" (plus a streaming variant) via the Gemini API, Google AI Studio, and in private preview on the Gemini Enterprise Agent Platform.
Momentum trend
03.05.01.08.
Features
| Compliance/Certification | Per Google, its safest robotics model to date; high compliance in safety-constraint and human-proximity benchmarks (ASIMOV-Agentic benchmark) |
| Deployment Model | Cloud API access (no required special hardware/software); streaming variant for latency-sensitive live robotics applications |
| Use Case Scope | Robotics: spatial reasoning, video understanding (progress/success detection), instrument reading, multi-robot orchestration, multi-step task planning |
| Integrations | Gemini Live API (bidirectional streaming), Google Search, function calling, code execution, VLA models and robot APIs (e.g. Boston Dynamics Spot) |
| License | Proprietary model, used via API under Gemini API Additional Terms of Service or Google Cloud Platform Terms of Service |
| Platform | Gemini API, Google AI Studio; private preview on Gemini Enterprise Agent Platform |
| Price | $2.00 per 1M tokens (text/standard), image input approx. $0.0011/image; no official pricing for enterprise access |
| Release Date | July 30, 2026 (public preview) |