Meta and Nvidia release Llama 3.3 Nemotron Ultra – a 253B open model
Meta and Nvidia have jointly launched Llama 3.3 Nemotron Ultra, a large language model (LLM) featuring 253 billion parameters. The model is open and designed for demanding reasoning tasks, challenging existing closed models.
What happened?
Meta and Nvidia have jointly published Llama 3.3 Nemotron Ultra, an open language model based on the Llama 3.3 architecture. The model features 253 billion parameters and has been optimised using Nvidia's Nemotron training process. It is intended to compete with leading proprietary models such as GPT-4o and Claude 3.5 Sonnet, particularly in tasks requiring advanced reasoning capabilities. The model is fully downloadable and licensed for self-hosting, giving users complete control over data and the operational environment.
Key facts
| Modellnamn | Llama 3.3 Nemotron Ultra |
|---|---|
| Antal parametrar | 253 miljarder |
| Modelltyp | Öppen vikt (open-weight) |
| Utvecklare | Meta och Nvidia |
| Träningsprocess | Optimerad av Nvidias Nemotron |
”Meta and NVIDIA have co-released Llama 3.3 Nemotron Ultra, a 253B open-weight model built on the Llama 3.3 architecture and optimized by NVIDIA's Nemotron training pipeline.”
”The model is fully downloadable and licensed for self-hosting, positioning it directly against frontier closed models like GPT-4o and Claude 3.5 Sonnet on reasoning-heavy tasks.”
Why it matters
The launch marks a significant development in the AI field by offering an open model capable of handling complex reasoning. This allows organisations to host a state-of-the-art model themselves, addressing needs for data sovereignty, low latency, and cost control. The collaboration between Meta and Nvidia, where Nvidia contributed its Nemotron post-training stack and Meta provided the base architecture, is also noteworthy.
Who is affected?
Developers, researchers, and enterprises working with advanced AI and machine learning are directly impacted. This particularly concerns those who require high-performance models but do not wish to rely on third-party APIs for sensitive data. End-users of AI-based products may benefit indirectly from improved performance in applications.
Impact on the EU
The model's openness and the possibility of self-hosting mean that EU-based organisations can operate it within EU borders, potentially making it easier to meet data protection requirements such as GDPR. Availability for EU users is immediate, as the model is downloadable globally.
What else you should know
Running Llama 3.3 Nemotron Ultra requires significant agricultural resources given the model's size of 253B parameters. This is a marked difference from the smaller Llama 3 model with 70 billion parameters.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Vilka bolag berörs?
The link opens in a new window and leads to the publisher's own site.
Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.
AI-verktyg i artikeln
Topics
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Decide whether this affects strategy over 6–12 months or is just noise.
- Discuss with leadership: do we own the right question or does ownership need to move?
- Ask: what risk are we taking by NOT acting on this this quarter?
Generated angle — not editorial analysis of "Meta and Nvidia release Llama 3.3 Nemotron Ultra – a 253B op"