Google/DeepMind

Latest AI news, models and releases from Google/DeepMind. ['A2UI v0.9', 'AI Co-Scientist', 'AI Evaluator', 'AI Mode', 'AI Overviews', 'Alexa', 'AlphaEvolve', 'AlphaFold', 'AlphaGenome', 'AlphaQubit', 'Antigravity', 'Auto frame', 'BERT', 'BioBERT', 'Chinchilla', 'ClinicalBERT', 'Computational Discovery', 'Confidential GKE Nodes', 'Co-Scientist', 'Empirical Research Assistance', 'Era', 'ERA', 'Executive LLM', 'Farmscapes 2020', 'Flood Hub', 'frontier AI', 'Frozen v2', 'FunctionGemma', 'Gemini', 'Gemini 1.5 Flash', 'Gemini-1.5-pro', 'Gemini 1.5 Pro', 'Gemini 2.0 Flash', 'Gemini 2.5', 'Gemini 2.5 Flash', 'Gemini-2.5 Flash', 'Gemini-2.5-Flash', 'gemini-2.5-flash-image', 'Gemini 2.5 Flash Lite', 'Gemini 2.5 Pro']

Official site ↗

Models 🇺🇸 ⚡

From GPT-2 to gpt-oss: Analyzing Architectural Advances and Comparison with Qwen3

OpenAI released gpt-oss-120b and gpt-oss-20b, their first open-weight models since GPT-2. The architecture features Mixture-of-Experts, Grouped Query Attention, RoPE, SwiGLU, and MXFP4 optimization for local inference. Comparisons with GPT-2 and Qwen3 highlight advances in width vs depth trade-offs and attention sinks.

OpenAI

Alibaba/Qwen

Automatic Blame Attribution in LLM-Based Multi-Agent Systems: New Research

Researchers from Penn State University and Duke University, in collaboration with Google DeepMind, introduced the task of automatic fault attribution in LLM-based multi-agent systems. They developed the Who&When benchmark with 127 fault logs and three attribution methods (All-at-Once, Step-by-Step, Binary Search). Experiments showed that even the best models (GPT-4o, o1, DeepSeek R1) perform poorly: accuracy in identifying the responsible agent is about 53.5%, and the error step is only 14.2%.

Google/DeepMind

Google DeepMind

OpenAI

DeepSeek

Synced27.07 · 17:05

Show more ↓