Reflection AI introduced Beam on October 5, a 501-billion-parameter open-weight sparse mixture-of-experts model with 23 billion active parameters that the Nvidia-backed startup says matches Z.ai's GLM…
Reflection introduced Beam, its first open-weight model, a sparse Mixture-of-Experts system with 501 billion total parameters and 23 billion active parameters built for coding, reasoning, and agentic …
OpenAI disclosed that it is reviewing 50 petabytes of historical training and evaluation records to audit autonomous agent intrusions, assigning roughly 7,000 GB200 and GB300 GPUs to an inspection pip…
DeepSeek CEO Liang Wenfeng plans to deploy at least 160,000 Huawei Ascend 950DT chips at a new data center in Inner Mongolia, with initial deliveries expected between Q4 2026 and Q1 2027, as the Chine…
SGLang and Miles added day-0 support for DeepSeek-V4.1, whose Engram memory component holds 189 GiB of fp8 weights across two tables. In paired tests on 4x GB300 (TP4/EP4), host offload of the Engram …
OpenAI unveiled Jalapeño, its debut AI accelerator chip, on 25 August, claiming up to 13.4 petaflops of 4-bit compute, 232 gigabytes of memory, 15.4 terabytes per second of memory bandwidth, and up to…
By the end of 2028, two frontier AI labs are on track to control most of the world's usable compute, according to Dylan Patel, founder of SemiAnalysis, who calculated that their compute grows 3x annua…
OpenAI unveiled benchmark results for its Jalapeño inference chip at Hot Chips 2026, claiming 1.7x to 3.6x lower end-to-end latency and 1.5x to 1.9x better work per watt than Nvidia's GB300 at half th…
OpenAI's first in-house AI chip, Jalapeño, outperformed Nvidia's GB300 in inference efficiency and speed per unit of power in initial benchmarks, signaling a growing challenge to Nvidia's dominance in…
At the 37th Hot Chips conference, OpenAI unveiled its custom inference chip Jalapeño, claiming it outperforms NVIDIA's GB200/GB300 systems with 1.5–1.9× more work per watt at peak throughput, 1.7–3.6×…
OpenAI has published benchmark results for Jalapeño, its first custom AI inference chip, claiming it outperforms NVIDIA's GB200 and GB300 in throughput, latency, and power efficiency simultaneously. T…
OpenAI unveiled Jalapeño, an in-house inference ASIC and system built with Broadcom, at Hot Chips 2026, claiming it outperforms NVIDIA GB200 and GB300 in throughput per kilowatt and latency for OpenAI…
OpenAI disclosed benchmark results for its custom AI inference chip, codenamed Jalapeño, showing it outperforms Nvidia's GB300 processors with 1.5 to 1.9 times more AI work per watt and 1.7 to 3.6 tim…
Nvidia's GB300 rack draws over 180 kilowatts, equivalent to the electricity load of 60 average American homes, and US data centers consumed approximately 176 terawatt-hours (TWh) in 2025, about 4.4% o…
A developer detailed deployment recipes for Qwen3.8-2.4T-A95B, a 2.4-trillion-parameter Mixture-of-Experts model with 95B active parameters, using vLLM. The guide covers verified GPU configurations, i…
DeepSeek released DeepSeek-V4-Pro-0813, the production successor to its V4 Pro preview, under an MIT license with open weights, adding a DSpark speculative decoding module and stronger agentic benchma…
StorageReview's lab-tested leaderboard for best local AI desktops in 2026 ranks the NVIDIA DGX Spark as the best overall deskside AI system, featuring a GB10 Grace Blackwell superchip with 128GB unifi…
Microsoft has formally accepted Horizon 1, the first phase of IREN Limited's AI cloud deployment under a five-year contract worth approximately $9.7 billion, with IREN on track to deliver 50 MW of IT …
Goldman Sachs estimates that memory components account for roughly 62% of the total material costs for Nvidia's upcoming Vera Rubin superchip, up from about 53% on the previous-generation GB300 system…
AI coding agents are eroding Nvidia's CUDA software moat by lowering the expertise barrier for GPU kernel development, according to a tech analysis. Tools like Anthropic's Claude Code can generate per…