NVIDIA hands OpenShell to the CNCF, and its agent policies move to Cedar
OpenShell, NVIDIA's runtime that enforces what AI agents may touch, is now a CNCF Sandbox project. It moves to a neutral ai-openshell GitHub org.
Topic · 35 stories · page 1 of 2
35 stories tagged nvidia on AI Tamer, newest first.
OpenShell, NVIDIA's runtime that enforces what AI agents may touch, is now a CNCF Sandbox project. It moves to a neutral ai-openshell GitHub org.
GR00T 2 learns robot skills from 400,000+ hours of human video. NVIDIA reports 40% zero-shot on unseen tasks, 95% after five demos. Code not out yet.
NVIDIA posts four Cosmos 3 Nano open-weight checkpoints for interactive driving and robot world simulation under OpenMDW1.1.
SkillSpector checks Claude Code, Codex and MCP skills for injection, exfiltration and supply-chain risk, then scores them 0 to 100. How to run it.
Vade runs Qwen3.8 Flash-Next on one to four DGX Sparks, writes its own GPU kernels in Rust and promises the same token IDs on every run.
cuda-oxide 0.3.0 drops the pinned nightly, shares one host runtime with cutile-rs, and previews CuTe in Rust. What changed and what is still missing.
Nemotron 3.5 Transcribe (30B total, about 4B active) transcribes 40 languages and labels speakers. NVIDIA also opened the Granary v2 data pipeline.
Nvidia says it will invest $3 billion in Meta's equity offering, priced at $625 a share, and help train and serve Meta's Muse models on Vera Rubin.
An OpenShell gateway with no auth configured gave every caller admin rights. OpenShell v0.1.6, out 27 October, makes it refuse to start unless you opt in.
Open Linux modules lift GeForce FP16 and BF16 tensor issue rates toward 2x on RTX 30 and 40 cards. Owners report 14 to 36% faster LLM prefill, not decode.
Dynamo 1.6.0 adds pluggable router queues, multi-replica Rust EPP and coding-agent session routing. One default change touches Claude Code traffic.
One developer rewrote NVIDIA's streaming RE-USE model in Rust and measured 8.4 ms per 20 ms frame on an M3 CPU, against 16.9 ms through WebGPU.
NVIDIA's Kumo Relational reads linked tables and predicts from up to 10,000 examples in context. Open weights under OpenMDW; benchmarks are NVIDIA's.
NVIDIA's Flare, now Hyperflare, is an Apache-2.0 local inference server for SM120/SM121 GPUs. Python prepares each model; a Rust runtime serves it.
Flora's segmenter, captioner and rewarder are on Hugging Face under a noncommercial license. NVIDIA's ~125x cost edge rests on an assumed GPU price.
Five open NVIDIA retrieval models search page images and text, rerank results and parse PDFs, under the permissive OpenMDW license. What to check first.
Modal's first Vera CPU server matched x86 on quiet single tasks but kept its pace as sandboxes piled up, doubling throughput in Modal's own early benchmark.
DeepInfer is Apache-2.0 and runs V4.1-Flash on NVIDIA SM100 or Huawei Ascend 950. This release is offline inference only, without KV cache or load balancing.
NVIDIA App 11.0.9.565 runs a local MCP server so AI agents can launch games and change settings. VideoCardz found it on by default. One toggle disables it.
NVIDIA's 16 October cards publish Dory 1.0 4B, a looped model with a 256K context, in BF16, MXFP8, NVFP4, and GGUF.
NVIDIA's 15 October posts release open-weight Nemotron 3.5 Super VL and Bolt. The 5x throughput line and partner figures are NVIDIA's.
NVIDIA's 13 October post says Run:ai is open source under DSX. The platform and its component repositories are under Apache 2.0.
NVIDIA and Tri Dao's lab describe ReplaySSM, which caches recent SSM inputs instead of writing full state every step. The speedups are the authors' benchmarks.
At the OCP summit in San Jose, NVIDIA and AMD each described a rack for AI agents, apart from GPU inference. Both are designs and plans, not shipping tests.