NVIDIA AI AI News, Models and Product Updates
Track latest Nvidia AI AI news, launches, research, and ecosystem moves.
Nvidia Ai news, model releases, product launches, research updates, and major announcements in one place.
Latest Articles
- Creating Power Flexibility in Datacenters via Query Routing — A query-routing framework lets LLM datacenters cut power draw on demand without deferring requests, reaching up to 16% power flexibility in a simulated 10,000-H100 site, with 90% of the target cut landing within about 10 seconds.
- Coarse Semantic Blocks for Efficient Training-Free Sparse Attention — A training-free sparse attention method that clusters keys into large semantic blocks delivers 2.3-2.4x prefill and 1.7-2.0x decode speedups at 512K tokens on Qwen3.5 models while staying close to dense accuracy for the 35B-A3B model.
- FP4Quant: Efficient 4-Bit Communication Compression for Model-Parallel LLM Training — FP4Quant compresses inter-GPU communication in model-parallel LLM training to 4-bit, reporting up to 1.56x end-to-end throughput over BF16 on H100s with Qwen-7B, with a trade-off in model quality.
- Formalizing PTX in Lean to Prove GPU Kernel Correctness — Researchers formalized NVIDIA's PTX in Lean and used it to prove correctness of simple WGMMA GEMM, FlashAttention and a three-kernel MoE pipeline, bringing formal proofs to deep-learning GPU kernels.
- GC-MoE: Overlapping Expert-Parallel Token Dispatch with FP8/FP4 Grouped GEMM via CUDA Green Contexts — GC-MoE uses CUDA Green Contexts to split GPUs into separate communication and compute partitions for MoE inference, speeding DeepSeek-V4.1-Flash decode by about 1.19x on 8 H20s and beating DeepGEMM Mega MoE.
- Asymptotic Optimality of $c\mu$-Rule for LLM Inference Scheduling under Memory Constraint — A queueing-theory paper shows a c-mu priority rule, with service rate set by compute, memory bandwidth and KV-cache capacity, is asymptotically optimal for LLM inference scheduling and cuts latency on an H800 serving engine.
- NVIDIA Shows 3 Ways to Speed Up MoE Inference on DGX Spark, Lifting Ling-3.0-flash Throughput 42.5% — NVIDIA showed three stacked optimizations that raised GSM8K decode throughput 42.5% for Ant Group's 124B Ling-3.0-flash MoE model on a single DGX Spark, with no detectable accuracy change.
- AI inference startup Fal doubles its valuation to $16B as demand surges — Fal raised $1.3 billion led by Bessemer at a $16 billion valuation, double its April mark, as demand for hosted inference on open-source video, image and audio models surges.
- AI agents take over at SETI's Allen Telescope Array, scanning for extraterrestrial intelligence — The SETI Institute and Breakthrough Listen now use AI agents on Nvidia Holoscan to steer the Allen Telescope Array and analyze radio signals live, cutting weeks of work to hours.
- AI Is Changing How Semiconductor Fabs Work — and Engineers Are Moving Faster Because of It — NVIDIA lays out its 'AI-defined fab' pitch, citing COMSOL and Applied Materials simulations running up to 10x faster with cuDSS as chipmakers face rising complexity and slow physical trials.