Nemotron 3 Ultra
Nemotron 3 UltraRankingin thinking mode?
Alternatives?
CheaperClaude Haiku 5.5max+50%-80%$0.21Remove sampling parameters+5
OpenGLM 5.3max+65%+115%$2.01Memory (4-bit): 336 GB → 452 GB+4
CheaperQwen3.6 35B A3BnoneOpen+10%-44%$0.25 / $1.49Context: 1M → 262K
CheaperGPT-5.6 Lunamax+5%-55%$0.18CheaperQwen 3.5 Flashunstatedsame-89%$0.065 / $0.26Claude Haiku 5.5max+50%-80%$0.21Remove sampling parameters+5Qwen3.6 35B A3BnoneOpen+10%-44%$0.25 / $1.49Context: 1M → 262KGPT-5.6 Lunamax+5%-55%$0.18Qwen 3.5 Flashunstatedsame-89%$0.065 / $0.26
Gemini 3.5 Flash Litelow-5%-15%$0.30 / $2.50Claude Haiku 5.5max+50%-80%$0.21Remove sampling parameters+5GPT-5.6 Sollow+30%+700%$0.26
DeepSeek V4 Pro 0813maxOpen×2.2+98%$0.67Memory (4-bit): 336 GB → 990 GB+5GPT-6.1 Solmax×4+300%$0.72Remove unsupported parameters+5GPT-6 Solmax×2.8+300%$1.04GLM 5.3maxOpen+65%+115%$2.01Memory (4-bit): 336 GB → 452 GB+4Qwen3.6 35B A3BnoneOpen+10%-44%$0.25 / $1.49Context: 1M → 262KGLM 5.2lowOpen-5%+115%$1.40 / $4.40Memory (4-bit): 336 GB → 452 GB
Kimi K2.6unstatedOpen-10%+71%$0.95 / $4Context: 1M → 256K+1Qwen3.5 397B A17BnoneOpen-10%+35%$0.60 / $3.60Context: 1M → 262KNemotron 3.5 Super VLOpenn/a—$0.19Nemotron 3.5 Bolt 74B-A9BOpenn/a——Context: 1M → 256KNVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4Openn/a—$0.08 / $0.45Memory (4-bit): not verifiedNemotron 3.5 LightningthinkingOpenn/a—$0.079NVIDIA Nemotron 3.5 Lightning 30B A3B NVFP4Openn/a——NVIDIA Nemotron 3 Ultra 550B A55B NVFP4Openn/a——Context: 1M → 262KNemotron 3.5 Content SafetyOpenn/a—$0.20 / $0.20Context: 1M → 128K+1Nemotron 3 Nano Omni 30B A3B ReasoningOpenn/a—$0 / $0Context: 1M → 256KAll NVIDIA models →Run it yourself
estimateBenchmarks
all categoriesFAQ
How much does Nemotron 3 Ultra cost?›
$0.50 per 1M input tokens and $2.50 per 1M output tokens; cached input $0.15 per 1M. A typical benchmark task costs about $0.58 at default effort.
Is Nemotron 3 Ultra free to use?›
There is no free API; the cheapest paid price is $0.50 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Nemotron 3 Ultra, and what is it best at?›
#90 of 334 in Chat & Knowledge, #63 of 132 in Coding, #92 of 133 in Agents & Tool Use, #3 of 245 in Writing & Language, #87 of 133 in Vision & Documents. Strongest at: Science, Long context, Knowledge, Instructions.
Is Nemotron 3 Ultra better than GPT-5.6 Luna?›
On Mystery Game Puzzles, Nemotron 3 Ultra scores 20 vs 21 — lower. It is 224% more expensive per task.
How can I use Nemotron 3 Ultra?›
Through the API of BaseTen, DeepInfra, Venice.
What is the context window of Nemotron 3 Ultra?›
1M tokens, with up to 66K output tokens.
When was Nemotron 3 Ultra released? Is there a newer version?›
Released 2026-06-04. It is the current version in its line.
Is Nemotron 3 Ultra open source?›
Yes, the weights are open (openmdw-1.1 licence).
Can I run Nemotron 3 Ultra locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Nemotron 3 Ultra read images?›
No, it accepts text only.
What is Nemotron 3 Ultra?›
Nemotron 3 Ultra is an open-weights language model from NVIDIA, released Jun 2026.
What are the best alternatives to Nemotron 3 Ultra?›
Claude Haiku 5.5 (cheaper), GLM 5.3 (open weights).