NVIDIA

Nemotron 3 Ultra

Nemotron 3 Ultra is an open-weight model from NVIDIA, released Jun 4, 2026. Its best result is #24 of 143 on LiveCodeBench with 85.98%. It costs $0.60 per million input tokens and $2.40 per million output tokens, with a 256K-token context window.

Developer
NVIDIA
Released
Jun 4, 2026
Context window
256K
Price in / out, per 1M tokens
$0.60 / $2.40
Weights
Open
Inputs
Text
Reasoning
Yes
Tool use
Yes
Hugging Face downloads, 30 days
1.1M
Hugging Face likes
371

Benchmarks

BenchmarkScore barAccuracyRankCost / test
SWE-benchCoding69.00%65 / 85N/A
Vibe Code BenchCoding7.64%99 / 115N/A
LiveCodeBenchCoding85.98%24 / 143N/A
Legal Research BenchLaw15.38%70 / 85N/A
MedCodeMedicine38.62%71 / 107N/A
GPQA DiamondReasoning86.11%44 / 138N/A
MMLU-ProReasoning85.76%55 / 138N/A

Vals AI, as of Oct 24, 2026.

Compare Nemotron 3 Ultra

NVIDIA

The GPU maker also publishes Nemotron, open models tuned for enterprise and reasoning. Founded 1993, based in Santa Clara. nvidia.com

Questions

Who makes Nemotron 3 Ultra?

Nemotron 3 Ultra is made by NVIDIA, founded 1993 and based in Santa Clara.

How much does Nemotron 3 Ultra cost?

$0.60 per million input tokens and $2.40 per million output tokens, at its provider’s list price.

What is Nemotron 3 Ultra’s context window?

256K tokens (262,144).

Is Nemotron 3 Ultra open source?

Nemotron 3 Ultra has open weights, downloadable from Hugging Face (nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16).

Sources

Every number on this page comes from these sources; models without a published number show none. Download the data. Cite as “Nexus Explorer, nexus.inc/models”.