News
NVIDIA released Nemotron 3.5 Super VL and two Bolt models to span data center, desktop and edge agent deployments.
NVIDIA said on Oct. 15, 2026, that it has expanded its Nemotron family with three open models aimed at agent deployments across servers, desktops and edge devices. The lineup is led by Nemotron 3.5 Super VL, a 120-billion-parameter model the company says is designed for specialized always-on agents, alongside two smaller Bolt models for local and edge workloads.
What NVIDIA released
According to NVIDIA’s announcement, Nemotron 3.5 Super VL is available now, as is Nemotron 3.5 Bolt 74B-A9B 24GB. Nemotron 3.5 Bolt E4B 4GB is scheduled to arrive on Oct. 23. NVIDIA said the three models are intended to cover a range of agent deployments, from enterprise infrastructure to local hardware, without forcing users to choose between capability and deployability.
The company described Super VL as its fastest open agentic model in its class, saying it delivers high accuracy on multimodal tasks such as instruction following, tool use, multistep business workflows and visual understanding. NVIDIA also said the model is built in collaboration with the Nemotron Coalition and performs at roughly one-fifth the parameter count of its largest reasoning model while targeting comparable accuracy in its class. NVIDIA cited Artificial Analysis Intelligence Index leaderboard results showing up to 5x faster throughput versus other models in its class.
Edge-focused Bolt models
Nemotron 3.5 Bolt 74B-A9B 24GB is positioned for local agentic workloads on NVIDIA RTX and DGX Spark systems. NVIDIA said it uses native 2-bit quantization to fit within 24GB of memory and to run alongside other applications. The company said the model is optimized to balance intelligence and speed per gigabyte, and that it can be used for local tasks such as web research, writing and code debugging without sending those workloads to the cloud.
Nemotron 3.5 Bolt E4B 4GB is aimed at physical and conversational agents on GeForce RTX, Jetson and RTX/DGX Spark hardware. NVIDIA said it is designed to combine natural conversation with multi-turn agentic capabilities for fast local voice assistants, interactive characters and embedded interfaces.
Open model strategy and deployment claims
NVIDIA framed the release as part of a broader open-model strategy for enterprise and developer deployments. The company said Super VL ships with open weights, open datasets and training recipes, and that organizations can post-train it on their own data and deploy it on their own infrastructure. NVIDIA also said the model can be run wherever data-residency and security requirements dictate.
The announcement said Super VL was trained on about 26 trillion tokens across pretraining and continued pretraining, then post-trained with roughly 3.8 million samples and reinforcement learning across dozens of environments. NVIDIA said it is releasing open datasets on Hugging Face covering agentic tool use, finance, instruction following, long context, fairness and vision-language understanding, along with reinforcement learning environments in NeMo Gym.
For production use, NVIDIA said the model is available through Hugging Face, Open Router, ModelScope community and as NVIDIA NIM. It also said the model runs across DGX Spark, DGX Station, RTX PRO and RTX systems, Jetson, and data center and cloud environments. NVIDIA said NIM provides a production-ready microservice with validated configurations and security and performance updates every two weeks.
Context and remaining limits
The release follows earlier Nemotron updates this year. Background material supplied with the announcement notes that NVIDIA introduced Nemotron 3.5 Lightning on Aug. 11, 2026, and that its Nemotron 3 family has also included Super and Ultra variants. The supplied background also says a separate partner announcement on Oct. 15, 2026, identified Nemotron 3.5 Super VL as the foundation for Multiverse Computing’s Comet 60B, underscoring ongoing ecosystem use around the model family.
The announcement does not provide independent market data, pricing or customer adoption figures beyond the examples named by NVIDIA, and it does not quantify the commercial impact of the release. The next verified date in the supplied material is Oct. 23, 2026, when NVIDIA said Nemotron 3.5 Bolt E4B 4GB is due to become available.
Be the first to comment