Tag: machine-learning
ATLAS: New Scaling Laws for Multilingual AI Models
Google DeepMind's Atlas research explores multilingual scaling laws, providing a framework for training highly efficient models across hundreds of languages.
NVIDIA Earth-2: World’s First Open AI Weather Models
NVIDIA launches Earth-2, a family of open, accelerated models for global weather forecasting, nowcasting, and data assimilation.
VibeVoice-ASR: Long-Form Speech Breakthrough
Discover VibeVoice-ASR, Microsoft's new model capable of 60-minute single-pass speech-to-text with speaker diarization and custom hotwords.
Liquid AI LFM2.5-1.2B-Thinking: Compact Power
Exploring Liquid AI's newest 1.2B reasoning model optimized for agentic tasks, RAG, and high-speed edge inference with LIV convolution architecture.
NVIDIA TTT-E2E: Test-Time Training Long Context
NVIDIA researchers introduce TTT-E2E, a test-time training approach that enables models to handle ultra-long contexts efficiently through dynamic adaptation.
Gemini 3 Flash: Frontier Intelligence for Speed
Google releases Gemini 3 Flash, a fast AI model with Pro-grade reasoning at Flash-level speed. Achieves 90.4% on GPQA Diamond and 3x faster than 2.5 Pro.
Meta Develops Mango and Avocado AI Models
Meta announces Mango image/video AI model and Avocado LLM, targeting first-half 2026 release. Led by Scale AI founder Alexandr Wang with $14B investment.
T5Gemma 2: Next-Gen Encoder-Decoder Models
Google DeepMind releases T5Gemma-2, a hybrid model combining the strengths of T5 and Gemma for industry-leading performance on translation and reasoning tasks.
Tencent HY 2.0: MoE Model with 73.4 IMO Score
Tencent releases HY 2.0 foundation model with MoE architecture, 256K context, and major gains in reasoning, coding, and instruction following capabilities.
UMA Launches Physical AI Robotics from Europe
UMA, founded by Tesla and Google DeepMind veterans, launches to build humanoid robots for real-world deployment in warehouses, hospitals, and factories.