[NVIDIA Releases Nemotron 3.5 Super VL and Bolt Open-Weight Models] According to Deep Tide TechFlow on October 19, as reported by TechCrunch, NVIDIA has launched a series of new open-weight models aimed at intelligent agents and edge computing. Among them, Nemotron 3.5 Super VL is a 120-billion-parameter model primarily designed for enterprise-level agents. NVIDIA claims that the model achieves accuracy in instruction-following and tool-usage tasks comparable to Nemotron 3 Ultra, and after fine-tuning for specific agent tasks, it can complete tasks faster than some models from Anthropic and OpenAI. NVIDIA also introduced two lightweight Bolt models optimized for local execution. Nemotron 3.5 Bolt 74B-A9B 24GB is a 74-billion-parameter model designed for local agent tasks such as multi-step reasoning, tool invocation, and web browsing. The smaller Bolt E4B is primarily intended for conversational agents. These models can run on AI PCs equipped with Nvidia RTX Spark chips as well as DGX Spark workstations. NVIDIA stated that Super VL and Bolt 74B-A9B are now available for use, while Bolt E4B will be released later this month. The company aims to encourage enterprises to deploy and train open models on their own infrastructure while reducing the need for users to send personal data and tasks to the cloud by enabling local agent execution.