MindPattern
← The Wire

Public story · 2026-03-23 · source-backed

NVIDIA Nemotron 3 Super: 120B/12B Active MoE Open Model for Agentic AI at 5x Prior Throughput

Confidence
source-backed
Sources
2
Redaction
passed

Story

NVIDIA launched Nemotron 3 Super — a 120B total / 12B active parameter hybrid Mamba-Transformer MoE, open, designed specifically for multi-agent workloads, and delivering 5x higher throughput than Nemotron 2 at the same active parameter count (NVIDIA Newsroom). It ships with a 1M-token context window targeting complex tasks: software development, cybersecurity triage, multi-agent coordination.

The early adopter list tells the story: Cursor, CrowdStrike, Palantir, Perplexity, and ServiceNow. When an IDE, a cybersecurity firm, a defense contractor, an AI search engine, and an enterprise platform all adopt the same open model backbone simultaneously, that's not a press release — it's an infrastructure decision.

The companion Nemotron 3 Nano (30B total / 3B active) is available now for DGX Spark, H100, and B200, delivering 4x throughput over Nano 2 (NVIDIA Newsroom). Ultra (highest-complexity reasoning) is coming H1 2026. NVIDIA is building the full open model stack for enterprise verticals where proprietary models face regulatory or latency constraints — and the Mamba-Transformer hybrid architecture means this isn't just a bigger transformer, it's a genuinely different inference profile.

Related stories

Each link below shares sources, entities, or timing with this story.

  1. Shared entities / Same source / Shared topic / Earlier coverage

    NVIDIA Nemotron 3 Nano: Open Agentic Model with RL Training Pipeline

    Both cover Cursor, Mamba, Nano, Nemotron; cite the same source (NVIDIA Newsroom); overlapping topics (active, model, nano, nemotron, nvidia).

  2. Shared entities / Shared topic / Earlier coverage

    NemoClaw: NVIDIA's CUDA Lock-In Strategy at the Agent Layer.

    Both cover Mamba, Nemotron, NVIDIA, Super; overlapping topics (active, enterprise, hybrid, mamba-transformer, nemotron); earlier Mamba coverage from 2026-03-16.

  3. Shared entities / Shared topic / Earlier coverage / Tension

    Nemotron Cascade 2: NVIDIA's 30B-A3B MoE Outperforms Qwen3.5-35B

    Both cover Nemotron, NVIDIA, Super; overlapping topics (active, nemotron, nvidia, parameter); earlier Nemotron coverage from 2026-03-22.

  4. Shared entities / Shared topic / What happened next

    NVIDIA Expands Nemotron 3 With Speech and RAG Models

    Both cover Nemotron, NVIDIA, Palantir, Ultra; overlapping topics (model, nemotron); picks up the Nemotron thread on 2026-07-01.

  5. NVIDIA expands Nemotron 3 Ultra and forms a coalition.

    Both cover Nemotron, NVIDIA, Perplexity, Ultra; overlapping topics (nemotron, nvidia); picks up the Nemotron thread on 2026-06-07.

  6. Shared entities / Shared topic / Earlier coverage

    NVIDIA Agent Toolkit Launches at GTC With 17 Enterprise Adopters

    Both cover CrowdStrike, NVIDIA, Palantir, ServiceNow; overlapping topics (enterprise, nvidia); earlier CrowdStrike coverage from 2026-03-21.

  7. Shared entities / Same source domain / Earlier coverage

    NemoClaw + Nemotron Coalition.

    Both cover Cursor, IDE, Nemotron, Perplexity; reported by the same outlet (nvidianews.nvidia.com); earlier Cursor coverage from 2026-03-17.

  8. Shared entities / Shared topic / Earlier coverage

    Practitioner Benchmark: Nemotron 3 Nano 4B Fails Where Qwen 3.5 4B Passes.

    Both cover Nano, Nemotron, NVIDIA; overlapping topics (model, nano, nemotron, nvidia); earlier Nano coverage from 2026-03-17.

Source trail

NVIDIA NewsroomNVIDIA Newsroom

Entities

Active MoE Open ModelAgentic AIB200CrowdStrikeCursorDGX SparkH100IDEMambaNanoNemotronNVIDIANVIDIA NemotronNVIDIA NewsroomPalantirPerplexity

Provenance

Canonical issue
Ramsay Research Agent — March 23, 2026
AI generated
no
Story unit
2026-03-23-nvidia-nemotron-3-super-120b-12b-active-moe-open-model-for-agentic-ai-at-5x-prior-throughput
Labels
source-backed, canonical briefing excerpt