/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Nvidia launches Nemotron 3 Ultra, a 550B-parameter MoE open model; Artificial Analysis says it is the smartest open US model, but trails Chinese model Kimi K2.6

It has roughly 550 billion total parameters, with about 55 billion active at any given time.

The Decoder Maximilian Schreiner

Context & Ripple Effects

Nvidia has been expanding Nemotron 3 from the family’s earlier 30B, 100B, and roughly 500B variants to a 120B Super model and a 30B-A3B multimodal Nano Omni model. Ultra extends that line at the high end while retaining the mixture-of-experts approach.

The family’s reported 50M-plus downloads gives Nvidia an existing distribution channel for open models. Artificial Analysis’ comparison matters because it places Nvidia at the top of the open US cohort while also showing that the leading edge remains internationally contested.

First-order effects

  • Developers seeking an open, US-built frontier-scale model gain a new Nvidia option with 550B total parameters but roughly 55B active per use, a design intended to concentrate computation on a subset of the model.
  • Nvidia strengthens Nemotron’s role as a model platform alongside its chips: Ultra gives the family a flagship counterpart to the smaller Super and multimodal Nano offerings.

Second-order effects

  • Other US open-model vendors face a clearer benchmark at the high end, while enterprise buyers can compare Nvidia’s model stack against Chinese alternatives rather than treating open-model choice as a single-market decision.
  • The active-parameter MoE design keeps inference efficiency central to model competition, reinforcing the value of deployment tooling and inference infrastructure around large open weights.

Third-order effects

  • If Nvidia continues pairing open models across sizes and modalities with its compute platform, competition may shift from selling a single frontier model toward controlling a full development-and-deployment stack.
  • The gap to Kimi K2.6 indicates that open-model leadership is likely to remain multipolar; rankings and practical deployment performance, rather than parameter totals alone, will determine which ecosystems gain adoption.

The trend: This is part of the shift toward open, mixture-of-experts model portfolios that compete on deployability and ecosystem reach as much as headline capability.

Discussion

  • @natolambert Nathan Lambert on x
    Nvidia is the tip of the spear for the growing US open model efforts. Is a long time coming, but obvious when a 550B model drops that more people will take notice. Many people won't realize the valuable training data they release too.