Google unveils Gemini Robotics 2.0 with enhanced dexterity and safety
Google's Gemini Robotics 2.0 can classify video frame completeness with 60 percent accuracy, marking a significant improvement over its predecessor.
Google's Gemini Robotics 2.0 can classify video frame completeness with 60 percent accuracy, marking a significant improvement over its predecessor.
Amazon Bedrock now offers explicit prompt caching for OpenAI GPT-5.6 models, with cached input tokens billed at a 90 percent discount.
OpenAI announced GPT-5.6 Luna and Terra with 80% and 20% lower prices, respectively, starting July 30, 2026.
Microsoft's AI division is focusing on small, specialized models that outperform rivals at half the cost, according to CEO Mustafa Suleyman.
Google DeepMind's Gemini Robotics 2 can control humanoid robots, enabling tasks like screwing in lightbulbs and tying trash bags, according to a recent release.
Hugging Face launches mDenseOn and mLateOn, two open 307M-parameter multilingual retrieval models trained on a 2.8B-pair translate-train corpus, achieving top results on BEIR and MIRACL benchmarks.
DeepMind released Gemini Robotics ER 2 on July 30, 2026, offering improved video understanding and task orchestration for robotics.
OpenAI reports GPT-5.6 Sol scores 38.3% on ARC-AGI-3 using two API settings, surpassing Opus 5's 30.2%.
OpenAI's GPT-5.6 Sol outperforms Claude Fable 5 on the Artificial Analysis Coding Agent Index at less than half the cost.
Pangram's new AI text detector, Pangram 4, makes one mistake per 24,000 documents in its own benchmarks, according to the company.
Google's Lyria 3.5 music model allows users to edit individual track sections without starting over, with tracks ranging from 30 seconds to 3 minutes.
DeepMind announced Lyria 3.5 on July 29, 2026, with improvements in musicality, lyrics, vocals, and creative control.