Public story · 2026-07-01 · high
NVIDIA Expands Nemotron 3 With Speech and RAG Models
Palantir also launched an engine to run Nemotron in air-gapped government systems, and Nemotron 3 Ultra has day-0 vLLM support.
Why now: NVIDIA dated the release June 30, covered here in the July 1 briefing.
Story
NVIDIA broadened its Nemotron 3 model family on June 30, adding open-weight models for speech and multimodal retrieval, per the NVIDIA Blog.
For builders who can't send data to a hosted API, this is the first concrete on-prem path. It bundles open weights, an inference server with day-0 support, and a sovereign hosting partner into one release.
Palantir launched an engine that runs Nemotron inside air-gapped US government environments. Agencies barred from cloud AI can deploy the same models NVIDIA ships publicly, without the data ever leaving their own network.
Nemotron 3 Ultra, the largest model in the family, got day-0 support in vLLM, the open-source inference server. That matters because the biggest model in a release is often the last one anyone can actually run. Here it's usable from day one.
The update also shipped open datasets and libraries aimed at agentic AI, plus new safety models. It's a release built for teams assembling their own stack rather than calling someone else's API.
A deployable stack, not a spec sheet
Open weights, vLLM, and sovereign hosting used to be three separate bets a team had to place and hope they'd line up. NVIDIA and Palantir wired them together into one deployable path. Regulated teams that couldn't touch a hosted API have somewhere to go. Palantir's air-gapped engine is the part that will pull sovereign AI work away from cloud vendors first, not last.
NVIDIA dated the release June 30, covered here in the July 1 briefing.
Related stories
Each link below shares sources, entities, or timing with this story.
Shared entities / Shared topic / Earlier coverage
NVIDIA Nemotron 3 Super: 120B/12B Active MoE Open Model for Agentic AI at 5x Prior Throughput
Both cover Nemotron, NVIDIA, Palantir, Ultra; overlapping topics (model, nemotron); earlier Nemotron coverage from 2026-03-23.
Shared entities / Same source domain / Earlier coverage / Tension
GPU vector indexing is now the default in OpenSearch Serverless, 10x faster at a quarter the cost.
Both cover NVIDIA, NVIDIA Blog, RAG; reported by the same outlet (blogs.nvidia.com); earlier NVIDIA coverage from 2026-06-25.
Shared entities / Same source / Shared topic / Earlier coverage
NVIDIA drops a massive open bundle: Nemotron models, 10T tokens, and an LLM Router blueprint.
Both cover Nemotron, NVIDIA; cite the same source (NVIDIA Blog); overlapping topics (model, nemotron).
Shared entities / Same source / Earlier coverage
NVIDIA drops a big open bundle for physical AI.
Both cover NVIDIA, NVIDIA Blog; cite the same source (NVIDIA Blog); earlier NVIDIA coverage from 2026-06-18.
Shared entities / Shared topic / Earlier coverage
Practitioner Benchmark: Nemotron 3 Nano 4B Fails Where Qwen 3.5 4B Passes.
Both cover Nemotron, NVIDIA; overlapping topics (agentic, model, nemotron); earlier Nemotron coverage from 2026-03-17.
NVIDIA Nemotron 3 Nano: Open Agentic Model with RL Training Pipeline
Both cover Nemotron, NVIDIA; overlapping topics (agentic, model, nemotron); earlier Nemotron coverage from 2026-02-25.
Shared entities / Earlier coverage
NVIDIA expands Nemotron 3 Ultra and forms a coalition.
Both cover Nemotron, NVIDIA, Ultra; earlier Nemotron coverage from 2026-06-07.
Shared entities / Shared topic / Earlier coverage
Nemotron-Cascade 2: 30B MoE Delivers Best-in-Class at 3B Active Params
Both cover Nemotron, NVIDIA; overlapping topics (agentic, model); earlier Nemotron coverage from 2026-03-20.
Source trail
Entities
Claim evidence
- NVIDIA Expands Nemotron 3 With Speech and RAG Models
Provenance
- Canonical issue
- 2026-07-01
- AI generated
- yes
- Story unit
- 2026-07-01-nvidia-broadened-nemotron-3-with-speech-multimodal-rag-and-safety-models-plus-a-palantir-soverei
- Labels
- source-backed, canonical briefing excerpt