NVIDIA is a semiconductor company best known for its GPUs, which are the primary compute hardware for training and running AI models. Its H100 and A100 chips power the majority of AI data centers worldwide.
Inferact, a startup whose team includes vLLM maintainers, released open-source "megakernels" for Google's TPU v7 (Ironwood) that delivered 709 tokens per second for a single user on Moonshot AI's Kimi…
Nvidia on Monday unveiled its Open Agent Safety Platform, an open-source security system the chipmaker says can stop AI agents from going rogue, with more than 100 companies using it at launch includi…
Nvidia released the Open Agent Safety Platform, a set of security protocols for enforcing safeguards on AI agents, days after CEO Jensen Huang called the safety warnings of Anthropic CEO Dario Amodei …
Nvidia announced OpenShell, an open-source runtime for securing autonomous AI agents, at GTC San Jose 2026 under the Apache 2.0 license. OpenShell enforces YAML-defined policies using kernel-level iso…
Nvidia released a set of AI agent safety tools on Monday that vice president and general manager of enterprise computing Justin Boitano said could have stopped this summer's Hugging Face breach if fro…
Nvidia released a software platform on September 28, 2026, aimed at preventing AI agents from misbehaving, according to CNBC Technology. Nvidia said the new software could have prevented OpenAI's Hugg…
Nvidia unveiled a new platform Monday to put guardrails on AI agents at both the software and hardware level, consisting of two parts: OpenShell and Sentry. OpenShell is open-source software that plac…
Nvidia is making its OpenShell agentic AI security sandbox generally available and introducing a new chip-level security platform called Sentry, which runs on Nvidia's Bluefield data processing units …
South Korean startup Xcena detailed its MX1 CXL Type 3 device at Hot Chips 2026, combining up to 2 TB of DDR5, SSD-backed InfiniteMemory capacity, and more than 1,000 custom RISC-V cores. Xcena chief …
Nvidia released Nemotron 3 Diarization, a roughly 100-million-parameter AI model that identifies which speaker is talking at any moment and can distinguish up to eight speakers, with weights freely av…
A developer built a private LLM chat client and server using MQTT as the transport, letting a phone reach an Ollama instance on a home GPU PC without opening any inbound ports. The system consists of …
Nvidia's RTX 5060 is usable for current PyTorch and image-generation stacks, but older packaged environments such as Fooocus's official portable still default to torch==2.1.0 and xformers==0.0.23, cau…
Nvidia's SoL-Pi system cuts coding agents' token usage by up to 49 percent with little change in performance by optimizing the control layer, or harness, between the model and its environment, accordi…
Nvidia unveiled NVQLink on October 28, 2025, an open architecture interconnect that links quantum processing units to Nvidia GPU-accelerated supercomputers with latency under 4 microseconds and throug…
Stanford and Nvidia released CLM-8B on September 23, an open-weights Contrastive Language Model that makes real-time agent decisions up to 9x faster than TypeSafe AI's proprietary Jev model, clocking …
An engineering analysis of GPU infrastructure economics finds that owning an 8x H100 SXM5-class server carries roughly $310,000 in upfront capital plus about $37,500 in electricity and $4,300 per year…
AMap, part of Alibaba Group, tested a small language model called Gwhere against its production multi-stage recommender for the "Guess Where You Go" homepage card, generating a likely next destination…
China Telecom AI released Xing4.0-29B-A4B on September 22, 2026, which the model card calls "the first model of this scale trained entirely on the Ascend NPU platform with the MindSpore framework," ru…
China Telecom AI released Xing4.0-29B-A4B, an open-weight agentic mixture-of-experts model under Apache 2.0 that was trained entirely on Huawei Ascend 910C NPUs with the MindSpore framework and no Nvi…
Cache reads, not output tokens, dominate the cost of running AI agents, according to an analysis of Nvidia's Open-SWE-Traces dataset priced with Fable 5 that found output tokens account for just 9-18%…