AI·SOURCE: Ars Technica·11 Kas 2026
DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
01 · EXECUTIVE SUMMARY & CONTEXT
The new model is up to twice as successful at some tasks.
03 · THE IN-DEPTH ANALYSIS
The new model is up to twice as successful at some tasks.