Nvidia ships Nemotron 3.5 models for enterprise and local agents
October 24: MarkTechPost reports Nemotron 3.5 Bolt E4B 4GB is released for local voice agents, game characters, and robots, without weights, license, or benchmarks.
Nemotron 3.5 Super VL, Bolt 74B, and the 4GB Bolt E4B are out, with related open models and local hardware.
Nvidia has released open-weight Nemotron 3.5 models for enterprise and on-device agents. Super VL, with 120 billion parameters, and Bolt 74B-A9B, with 74 billion parameters, were available as of October 19; MarkTechPost reported on October 24 that the smaller Bolt E4B, about 4GB, is released for local voice agents, game characters, and robots. Nvidia also announced Embed, Rerank, and Parse retrievers and a wider open portfolio including Kumo, Cosmos 3.5, Alpamayo 2 Edge, BioNeMo RFD4-Proteina, and Ising. At GTC Berlin it linked the models to RTX Spark, DGX Spark, and DGX Station and named European financial firms and other companies as users. Speed, accuracy, and cost figures are Nvidia or company claims.
- Super VL has 120 billion parameters; Bolt 74B-A9B has 74 billion and a 24GB variant.
- Bolt E4B is described as a 4GB model; no license or benchmarks were available.
- RTX Spark, released October 16, offers up to 128GB unified memory and 1 petaflop.
- DGX Spark 64GB systems were set to start shipping October 23; clusters of up to eight are supported.
- DGX Station offers 748GB coherent memory; a Windows version is expected in December.
- Named users include Revolut, Starling, Lloyds, bunq, NatWest, Modulr, Stripe, and others.
Timeline
dated events in the reporting · oldest first- Oct 16, 2026RTX Spark released for Windows personal agents · NVIDIA Blog
- Oct 19, 2026Nemotron 3.5 Super VL and Bolt 74B available · TechCrunch
- Oct 21, 2026Nvidia expands open models and retrieval lineup · NVIDIA Blog
- Oct 23, 2026DGX Spark 64GB systems begin rolling out · NVIDIA Blog
- Oct 24, 2026Bolt E4B 4GB reported released for local agents · MarkTechPost
What each outlet adds
beyond the shared facts- TechCrunchSuper VL and Bolt sizes, named users, and initial availability
- NVIDIA BlogWider model portfolio, hardware specs, and European financial deployments
- Hugging Face BlogNames Embed, Rerank, and Parse; body unavailable
- MarkTechPostReports Bolt E4B 4GB release without further specifications
Coverage
oldest first · one row per article · UTC- · 6d ago
TechCrunch · AIFirst report
Nvidia releases open Nemotron 3.5 models for agents and PCs
The 120-billion-parameter Super VL and 74-billion-parameter Bolt model are available now; Bolt E4B follows later this month.
- · 4d ago
NVIDIA Blog
NVIDIA announces open models for language, robotics, biology and quantum
The portfolio adds Nemotron 3.5, Kumo, Cosmos 3.5, Alpamayo 2 Edge, BioNeMo RFD4-Proteina and Ising.
- · 4d ago
Hugging Face Blog
NVIDIA announces Nemotron 3.5 Embed, Rerank, and Parse models
The post body was unavailable, so sizes, benchmarks, licenses, and release status are not stated.
- · 4d ago
NVIDIA details local AI hardware lineup at GTC Berlin
DGX Spark 64GB systems start shipping October 23, eight-unit clusters are supported, and Windows DGX Station is due in December.
- · 3d ago
European finance firms show NVIDIA-backed models at GTC Berlin
Revolut says PRAGMA starts with 80 million customers, and bunq claims 60% of closed-model cost per answer.
- · 1d ago
MarkTechPost
NVIDIA releases open Nemotron 3.5 Bolt E4B 4GB
The headline positions the small open model for local voice agents, game characters and robots.