NVIDIA launches Nemotron 3.5, Cosmos 3.5 and more open models for Europe's sovereign AI push
testingcatalog · x · 2026-10-21
NVIDIA announced a major expansion of its open model portfolio under a Europe-focused "sovereign intelligence" initiative:
- Nemotron 3.5 Super VL / Bolt: open language models supporting text and images for agent orchestration and reasoning, built with the Nemotron Coalition including Black Forest Labs, H-Company, and Mistral
- Nemotron Retriever: embedding, reranking, and document parsing models
- Kumo family: open-weight foundation models for structured business data
- Also Cosmos 3.5 world models, Alpamayo 2 Edge for autonomous vehicles, BioNeMo for healthcare, and Ising for quantum computing
Enterprises including Dassault Systèmes, SAP, Salesforce, Shopify, Palantir customers, and WPP are building post-trained applications on Nemotron. All models can be customized with local data and deployed on self-controlled infrastructure. The poster notes nothing especially exciting for general consumers so far, with new cars also joining NVIDIA's Hyperion AV network.
Related event: NVIDIA Unveils Nemotron 3.5 and Seven Open Model Families at GTC Berlin(5 posts)→
More from Models
- NatWest benchmarks British-accent ASR: Cohere Transcribe hits 15.3% WER — Nils_Reimers · 2026-10-21
- Building the same library twice: open-weight model cost $19.86 vs Opus's $346.70 — robeybeswick · 2026-10-21
- Pangram AI detector flags 2016 human-written papers as 100% AI-generated — GaryMarcus · 2026-10-21
- User's riddle stumps ChatGPT, Claude and friends — screenshot inside — DoItYourselfMate · 2026-10-21
- Grok voice mode shines but drains the $30 plan; ChatGPT talk mode lags — DevDminGod · 2026-10-21
- Independent security tests of Gemini 4 Argon diverge from Google's launch table — Thirumalaivasan_GJ · 2026-10-21