Tech

DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 10/11/2026, 15:28:23 UTC.

Briefings

Google DeepMind announced Gemini Robotics 2.5, a new version of its AI platform capable of performing complex physical tasks in real-world settings like Mercedes-Benz factories. The system uses a vision-language-action (VLA) model to interpret video inputs and generate precise motor commands, enabling dexterous actions such as tying a bow with a five-fingered hand. The platform includes over 200 evaluated activities, including serving food with tongs, and aims to operate alongside humans with an open-source safety benchmark.

Key points

• Gemini Robotics 2.5 uses a vision-language-action (VLA) model to interpret video inputs and generate precise motor commands for physical tasks.
• The system was evaluated across more than 200 activities, including serving food with tongs and using a pipette in laboratory settings.

Context

Google DeepMind's physical AI platform uses computational layers between user input and robotic action, including a vision-language model for planning and a VLA for generating actions. This release marks a significant step toward real-world task completion, with the system now capable of performing dexterous tasks in industrial environments.

Limitations

Evidence is limited to a single article from November 2026. Specific dates for the benchmark release or operational timelines are not provided. The evaluation scope of 200 activities is mentioned but does not specify the exact date or duration of testing.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 10/11/2026, 16:52:02 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeicudimh4oyopcvhzbvzjtdy72be76pup3i5l6l2mrcpzknzqqkvje