NVIDIA Nemotron 3 Super 120B-A12B

NVIDIA

Catalogue only Open weights

Sovereignty and hosting

Catalogue only

This model is catalogued for reference. It is not hosted on LLM EU yet, and no first-party endpoint serves it here. Requests to it go to the provider directly, under that provider's own terms.

Summary

Nemotron 3 Super is NVIDIA’s 123B mixture-of-experts model with about 12B active parameters, released under the NVIDIA Nemotron Open Model License. Its configuration allows 262,144 tokens and the model card advertises context up to 1M tokens, with RULER scores of 96.30 at 256k and 91.75 at 1M. The card lists English, French, Spanish, Italian, German, Japanese and Chinese as supported languages.

Specifications

Provider
NVIDIA (US)
Context window
262k
Parameters
123.61B
Modalities
text
License
other · licence
Open weights
yes
Release date
10 Mar 2026
Tasks
reasoning, coding, rag, summarization

Languages

Strong: en, fr, es, it, de

Adequate: ja, zh

Languages

Strong

  • en
  • fr
  • es
  • it
  • de

Adequate

  • ja
  • zh

Not verified: the provider has not documented this claim. — a language is listed as strong only where a source claims it; the card links the evidence where one exists.

When to use it

Strengths

  • strong published long-context scores up to 1M tokens
  • 12B active parameters keep inference cost below the total size
  • inference and agentic benchmark tables published with harness links
  • source-available vendor licence with published evaluation tooling

Limits

  • NVIDIA Nemotron Open Model License, a vendor licence rather than an OSI one
  • seven declared languages, fewer than the European-focused models
  • long-context scores come from NVIDIA’s own harness