NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE). Built on a hybrid Transformer-Mamba mixture-of-experts architecture, it...
With CoreAI, you can start chatting with NVIDIA: Nemotron 3 Ultra instantly — no separate subscription needed. CoreAI lets you chat with NVIDIA: Nemotron 3 Ultra alongside 300+ other AI models from NVIDIA and other providers like OpenAI, Anthropic, Google, Meta, and more.
NVIDIA: Nemotron 3 Ultra holds the #75 spot for context size out of 321 chat models in the CoreAI catalog — 512K tokens, roughly 384,216 words or 768 pages of text in a single conversation. On price, it is the #57 cheapest of 146 standard-tier models by input cost. It joined the catalog on 2026-06-04. It supports step-by-step reasoning (thinking mode).
Every model in this table is available on CoreAI — open the web app and run the same prompt through NVIDIA: Nemotron 3 Ultra and NVIDIA: Nemotron 3 Super side by side to judge with your own eyes.
A typical question (about 1,000 tokens in, 500 out) runs $0.00240 on NVIDIA: Nemotron 3 Ultra's raw per-token rates; a heavy session with a 50K-token document costs about $0.048. On CoreAI you don't juggle these numbers — one subscription covers NVIDIA: Nemotron 3 Ultra and 300+ other models under a plan usage limit, with no API keys. See plans.
NVIDIA: Nemotron 3 Ultra handles 512K tokens of context — roughly 384,216 words, or about 768 pages of text in one conversation.
NVIDIA: Nemotron 3 Ultra's underlying rates are $0.6 per 1M input tokens and $3.6 per 1M output tokens. On CoreAI it is included in one subscription alongside 300+ models, so there is no per-token billing to manage.
No — NVIDIA: Nemotron 3 Ultra is text-only. If you need image understanding, its standard-tier peers with vision are listed in the comparison table above.
Yes. NVIDIA: Nemotron 3 Ultra can work through problems step by step before answering; on CoreAI you can toggle thinking mode per message.
Open the CoreAI web app or the iOS/Android app, pick NVIDIA: Nemotron 3 Ultra in the model selector, and chat — no API key, no per-token billing. One subscription covers NVIDIA: Nemotron 3 Ultra and 300+ other models.
The closest alternatives on CoreAI are NVIDIA: Nemotron 3 Super, Thinking Machines: Inkling, inclusionAI: Ling-2.6-flash — similar capability. The fastest way to choose is running one prompt through them side by side in CoreAI's Compare view.
Chat with NVIDIA: Nemotron 3 Ultra and 300+ other AI models — all in one app.