Tag: Nvidia
An open model sorts eight overlapping speakers in real time
NVIDIA released Nemotron 3 Diarization, a small open model that tops VoiceArena's leaderboard while running live conversation streams.
NVIDIA teaches its robotics stack to take instructions from agents
The release adds agent-ready skills and ROS Lyrical support, alongside a pose library that runs up to 5.5…
SoL-Pi cuts coding agent token bills by rewriting the harness
The four mechanisms came out of automated research loops and trim token traffic by nearly half on a…
Nvidia starts certifying the power gear behind its AI factories
A new qualification program puts batteries and cooling units through Nvidia's DSX reference requirements.
Runway pitches live video you steer while it generates
The company wants video models to stream frame by frame instead of rendering after a prompt.
Huawei pulls Ascend 960DT launch forward to early 2027
Huawei has moved its next-generation Ascend AI chip up by two quarters as it tries to close the…
Nvidia’s Vera Rubin racks triple throughput in first MLPerf run
Nvidia's first MLPerf Inference submission for its Vera Rubin NVL72 rack reports up to 3.7x the throughput of…
CoreWeave scales Vera Rubin past a single rack for agents
CoreWeave has brought up multi-rack Vera Rubin NVL72 systems on its cloud, joining hundreds of Rubin GPUs into…
Salesforce trains its own reasoning model on Nvidia’s open weights
Koa is post-trained on Nemotron 3 Super with synthetic data, giving Agentforce a homegrown option for multi-step work.
Crusoe lands Perplexity for training and serving on GB300 racks
The multi-year deal puts Perplexity's full model lifecycle on Crusoe Cloud and hands enterprise AI tools to Crusoe's…
Nvidia pitches power efficiency as the new AI data center metric
At a Santa Clara event that doubled in size, Nvidia pitched a power optimization layer it says delivers…
NVIDIA open-sources OSMO to run robot pipelines from one file
NVIDIA has released a Kubernetes-native orchestrator that carries a robotics workflow from training cluster to factory floor.