All news

Google DeepMind releases Gemini Robotics 2.5 and open ASIMOV-Jailbreak safety benchmark

Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .

Language-model-driven robots completed real factory trials with 96% success, and alongside growing capabilities comes a common way to measure how hard they are to steer toward dangerous actions.

DeepMind has released a new version of its robot control system, Gemini Robotics 2.5. According to the company, task success rates rose 30–100% versus the previous version and speed by 20–30%, while the model is noticeably better at small, precise movements: the Apptronik Apollo 2 humanoid ties a bow, and the model handles serving food with tongs and working with a pipette. In trials outside the lab, Apollo 2 robots at Mercedes-Benz plants in Europe assembled auto parts with 96% success — which is still not enough for industrial use, DeepMind notes. The simulation environment for training was generated with Gemini 4 Argon, but the model itself runs on cloud servers, meaning there is latency. For safety testing, an open set called ASIMOV-Jailbreak was released: 1,500 prompts that provoke a robot into unsafe actions; with Argon in the loop, the system blocks 98.8% of attacks. The model is available to partners — Agile Robots, Apptronik and Boston Dynamics.

History

Release of Gemini 4 models: Argon and Carbon

11 events
  1. Gemini 4 Carbon leads MathArena, but solving a problem costs 15 times more than GPT-6.1 Sol
  2. Google releases Gemini 4 Hydrogen for fast agents and upgrades Search
  3. Google DeepMind releases Gemini Robotics 2.5 and open ASIMOV-Jailbreak safety benchmarkOpen now
  4. Google releases Gemini 4 Carbon to everyone and unveils /swarm mode — a team of agents from a single prompt
  5. Google unveils Gemini 4 Carbon, its strongest model, and opens it first to cyber defenders
  6. Four Gemini 4 Argon agents beat Minecraft by defeating the Ender Dragon in 27 hours 14 minutes, the paper's authors report
  7. Gemini 4 Argon safety report: the model considered hijacking a compiler to evade detection and backed off
  8. Gemini 4 Argon taught to control robots: pick-and-place success rose from 56% to 75% in Google's test
  9. Google releases Gemini 4 Argon, a new flagship model with a 1 million token context window
  10. Google is testing Gemini 4 Carbon: an employee compared its coding to Anthropic's best model, Opus 5.5
  11. Google releases Gemini 4 Argon: output up to a million tokens and first places in coding and enterprise benchmarks