Nemotron 3 Ultra
nvidia/nemotron-3-ultra-550b-a55b- NVIDIA
- Family · Nemotron 3 Ultra
- Open weights
- Text → Text
- Reasoning · adjustable
- Tools · JSON
Not on OSMEXO yet
Spec & list price
Context1M
Max output183K
Input / 1M$0.60
Output / 1M$2.404× input
Usage
Tokens / day440.6B▼ 8.4% 7d
Usage rank#12
Requests / day345.2K
Tool calls / req0.14
Per request · 30 days
Avg input10.1K0.7× median
Avg output5190.8× median
Cache hit70%about median
Reasoning share54%about median
Tool-call failures3.8%2.0× median
Consumption
Peak 9.6B 19 OctLow 1.6B 26 SepRank today #12
News & trends
14 mentions this week ▲ 367%this model · videos, posts, articles · 30 days
Trending · 14 days
- Nvidia’s Nemotron 3 Ultra: The Open AI Strategy Behind It #shorts816 views
- NVIDIA Nemotron 3 Ultra: бесплатная ИИ-модель для кодинга вместо Claude Code и Codex5.9K views
News · latest from NVIDIA
- Accelerate agentic AI with NVIDIA Nemotron models on Amazon EKS and SageMaker HyperPod EKSAWS · 29 Oct
- NVIDIA Open Models Advance Europe’s Sovereign Intelligence Across IndustriesNVIDIA · 21 Oct
- NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal ReasoningNVIDIA · 15 Oct
- Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt ModelsNVIDIA · 15 Oct
Speed
Output speed#170/33466 tok/sabout median
Fastest provider125 tok/sbest of 6 providers
First token#1661.02 sabout median
Technical details
- Reasoning
- on · high
- Tokenizer
- Other
- License
- other
- Parameters
- 561B
- Moderation
- none
- Downloads · likes
- 1.4M · 381
Reasoning effort
- high · default
- medium
- none
Request parameters 18
- tools
- tool_choice
- response_format
- structured_outputs
- reasoning
- reasoning_effort
- include_reasoning
- temperature
- top_p
- top_k
+8 more
- min_p
- max_tokens
- stop
- seed
- frequency_penalty
- logit_bias
- presence_penalty
- repetition_penalty
When to use
Best for
- Hard reasoning with adjustable effort
- Very long document analysis
- Enterprise agents with tool calling
- Self-hosted large-model deployments
Think twice for
- Image or audio input
- High-volume simple chat (cheaper options exist)
- Latency-critical interactive apps
Research papers
- Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI28 Oct
- Before the Rollout Ends: Early Terminal Reward Prediction for Long-horizon Coding Agents25 Sep
- Rethinking Heterogeneous System Disaggregation for Subquadratic Attention11 Sep
- An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics9 Sep
- Moral Hazard in Multi-Agent Language Models27 Jul
- Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning12 Jun
- Automating MD simulations for Proteins using Large language Models: NAMD-Agent10 Jul
Want Nemotron 3 Ultra on OSMEXO?
Sign in to request it. Requests are reviewed; a request does not guarantee availability.











