A developer's comparison of the RTX 5090 and RTX 4090 for AI workloads found that the 5090's jump to 32 GB of VRAM and 1,792 GB/s memory bandwidth is the most significant difference, not raw benchmark…
StepFun released Step 3.7 Flash, a 198B-parameter Mixture-of-Experts vision-language model with 11B active parameters, for enterprise deployment on NVIDIA-accelerated infrastructure. The multimodal AI…
Together AI built the world’s fastest speech-to-text stack, enabling NVIDIA’s Parakeet-TDT 0.6B v3 model to transcribe roughly 20 hours of speech in under 10 seconds. The company achieved this by opti…
NVIDIA Research presented eight new papers at the International Conference on Robotics and Automation (ICRA) demonstrating simulation-trained policies transferring to physical robots. The COMPASS navi…
The Kog AI team implemented a single-kernel LLM inference engine on AMD MI300X GPUs, achieving over 3,000 output tokens per second per request for a 2B-parameter model in FP16 precision. The monokerne…
Kog Team researchers introduced Delayed Tensor Parallelism (DTP), a Transformer architecture that hides communication overhead behind computation and weight streaming to accelerate batch-size-one infe…
Corsair Gaming launched a lineup of AI workstations and servers under the Corsair Pro brand on May 22, built around NVIDIA's Blackwell architecture and targeting enterprise customers. The flagship Fle…
Nvidia launched *007 First Light* on GeForce NOW, allowing subscribers to stream James Bond’s origin story from any device without downloads. The game is included with a limited-time purchase of a 12-…
SPAN announced the launch of XFRA, a distributed data center solution that places enterprise-grade, liquid-cooled compute nodes in residential and small commercial spaces to deliver gigawatts of new A…
Taiwan-based AI server manufacturer Wiwynn warned that critical component shortages for AI infrastructure will persist through 2028, extending beyond GPUs and memory to high-end PCBs, CPUs, and DRAM. …
Mistral AI secured $830 million in debt financing to build a data center near Paris, housing 13,800 NVIDIA Grace Blackwell GB300 GPUs with 44 MW of power capacity, with operations expected to begin in…
NVIDIA Vice President Bryan Catanzaro stated in April 2026 that AI compute costs now exceed employee payroll costs, a warning most enterprises ignored while focusing on "tokenmaxxing." Microsoft cance…
A new open-source coding agent, mini-SWE-agent, achieves up to 74% on the SWE-bench verified benchmark using just 100 lines of Python code. Developed by the Princeton and Stanford team behind SWE-benc…
China's April exports surged 14.1% year-over-year, nearly double the median forecast, driven largely by AI-related goods and integrated circuits that accounted for roughly half of the growth. Integrat…
NVIDIA's ASUS Ascent GX10 edge AI system, powered by the GB10 SoC, lacks any supported software interface for CPU energy measurement, making process-level energy attribution impossible on the platform…
NVIDIA's CUTLASS library, a header-only C++ template framework for writing custom CUDA kernels, powers much of the AI infrastructure behind FlashAttention, vLLM, and PyTorch's internal kernels. The li…
A compiler engineer at SemiAnalysis spent over $10,000 in a single afternoon running AI agents to fuzz NVIDIA's closed-source ptxas compiler, uncovering approximately 80 miscompilation bugs in three d…
AMD's ROCm (Radeon Open Compute platform) has reached production-ready maturity as an open-source alternative to NVIDIA's CUDA, now supporting LLM inference, fine-tuning, and image generation on AMD G…
AMD's ROCm platform now supports PyTorch, Ollama, LM Studio, and ComfyUI out of the box, enabling local AI workloads on AMD GPUs without the compatibility issues that plagued earlier versions. Users w…
NVIDIA has introduced Dynamo Snapshot, a checkpoint/restore approach for AI inference workloads on Kubernetes that reduces cold-start latency from minutes to near-instantaneous startup times. The solu…