Doubleword, one of six companies in the first wave of UK Sovereign AI investments, achieved up to 3× the throughput of vLLM for DeepSeek-V4-Flash on a single node of Isambard-AI, the UK's national AI …
Google has released Gemma 2, the next generation of its open models, featuring a 27B parameter version that offers performance competitive with models more than twice its size while being efficient en…
SpecLA, a speculative decoding runtime for stateful linear-attention models, achieves up to 1.70x end-to-end speedup over autoregressive decoding on an NVIDIA H100 with a GDN-1.3B target, according to…
Thinking Machines' open Inkling model, with 975 billion total parameters and 41 billion active, fits on a single high-memory box at 2-bit or 3-bit quantization, unlike the datacenter-scale Kimi K3. Re…
Meta Platforms is in advanced talks to lease computing power to Anthropic in a deal that could be worth $10 billion, according to a New York Times report. The agreement would make Meta a critical infr…
Google Cloud has released a blueprint for securing AI workloads on Google Kubernetes Engine (GKE), consolidating controls across multiple services to address infrastructure, model supply chain, and ap…
A developer released the first open-source training kernels for MiniMax Sparse Attention (MSA) on Hopper and Blackwell GPUs, enabling efficient million-token training with sparse attention. The kernel…
Google released Gemma 2, an open model family with 9B and 27B parameter sizes, featuring architectural changes like hybrid attention and Grouped-Query Attention for improved inference efficiency. The …
Crusoe launched Serverless Fine-Tuning and Self-Serve Deployments for its Intelligence Foundry, enabling organizations to fine-tune and deploy open AI models with granular control and cost efficiency.…
Ali, a 16-year-old independent researcher, developed PsiLogic, a chaos-aware optimizer built on Adam, and seeks an arXiv cs.LG endorser to submit his preprint. His benchmarks on an NVIDIA H100 GPU sho…
Google's Gemma 2 models demonstrate that architectural efficiency can deliver competitive performance with fewer parameters. The 27B model rivals models twice its size through hybrid attention, Groupe…
Recursive's automated AI research system achieved state-of-the-art results on three benchmarks: fixed-budget language model training, small-model training speed, and GPU kernel optimization. The syste…
Nick Creighton, an operator who ships, provides a detailed blueprint for deploying Meta's Llama 3 706B model locally, emphasizing privacy, latency, and cost benefits over cloud APIs. He outlines the e…
Google DeepMind released DiffusionGemma on June 10, 2026, a 26B open-weight text diffusion model that generates 256 tokens simultaneously, achieving up to 1,008 tokens per second on an H100—4-5x faste…
Google AI and Google DeepMind researchers released DiffusionGemma, an experimental 26-billion-parameter Mixture of Experts open model that uses text diffusion to generate entire blocks of text simulta…
Researchers at arXiv propose a hybrid pre-training objective for language models that combines a Joint Embedding Predictive Architecture (JEPA) latent-space prediction loss with masked language modell…