NVIDIA e Microsoft expandem IA acelerada do Azure para PCs Windows
A pré‑visualização Vera Rubin da Azure roda em hardware NVL72, escalando para 10,240 GPUs para treinamento DeepSeek‑V3 e oferecendo desempenho de inferência superior no DeepSeek‑R1, acco
Azure’s Vera Rubin preview runs on NVL72 hardware, scaling to 10,240 GPUs for DeepSeek‑V3 training and delivering top inference performance on DeepSeek‑R1, according to recent MLPerf results.
The companies also expanded Microsoft Foundry, offering serverless Programmed Training and Managed Compute for customizing NVIDIA Nemotron 3.5 models, with EY reporting a 55% accuracy gain and 15× cost reduction for a purchasing‑record classifier.
Local deployment options now include Azure Local support for RTX PRO 6000 Blackwell GPUs, Jetson Thor/Orin edge platforms, and RTX Spark‑based Windows PCs that can run large language models directly on the device.
A limited preview of Direct Stream for GPU‑accelerated Windows 365 Cloud PCs will enable up to two 4K streams at 60 fps, with a public preview slated for early 2027.
O que isso significa para você
Enterprises gain more flexibility to run high‑performance AI close to their data, reducing latency and reliance on cloud bandwidth. For users of free or locally hosted models, the RTX Spark rollout offers a path to experiment with large language models on consumer hardware, complementing the top‑ranked free chat models such as Kimi K3 and GLM‑5.3 listed on FreeLLMArena.
No FreeLLMArena
From our rankings on 21 Nov 2026, based on Arena's blind votes:
- Kimi K3 (Max): #1 among free chat models
Fontes
- NVIDIA and Microsoft Expand Accelerated Computing From Azure to Windows PCs, NVIDIA (official), 19 Nov 2026
Escrito automaticamente a partir das fontes acima, depois verificado pelo nosso código: todo número e nome deve aparecer nelas, e nenhuma frase pode copiar uma fonte. Encontrou um erro? Reporte‑o.
Qual IA você pode usar grátis hoje?
Todo modelo de IA gratuito, classificado por votos humanos cegos, com a memória e velocidade que ele precisa na sua própria GPU.