Nemotron 3 Ultra

nvidia/nemotron-3-ultra-550b-a55b

open reasoning model for long-horizon autonomous agents.

cmd --model nvidia/nemotron-3-ultra-550b-a55b
Intelligence index
38.3
Output speed
122.2 tok/s
Input
$0.60 /M
Output
$2.40 /M
Cache read
$0.12 /M
Agent-loop cost
$0.26 /M in
Context window
1M tokens
Released
June 4, 2026
Modalities

vs. the lineup

Nemotron 3 Ultra beside its stablemates and nearest rivals. The ◆ marks the best value in each column across every row shown.

pin a rival:
ModelIntelligenceCodingSpeedInput $/MOutput $/MBlended $/MContext
Muse Spark 1.2 Contributor56.872.2$0.10$0.20$0.131.05M
Qwen 3.7 Plus39.455.954.9$0.40$1.60$0.701M
MiniMax M2.738.952.6$0.30$1.20$0.52200K
Nemotron 3 Ultra38.349.3122.2$0.60$2.40$1.051M
MiMo V2.53856.877.3$0.14$0.28$0.181M
Step 3.7 Flash30.939.6391.3$0.20$1.15$0.44256K
DeepSeek V4 Pro (latest)pinned45.359.463.3$0.66$1.98$0.991M
DeepSeek V4 Flash (latest)pinned5269.1115.9$0.22$0.66$0.331M

coding performance

The Intelligence Index and its sub-scores, ranked against every scored model in the catalog. A metric that has not been measured for Nemotron 3 Ultra has been left empty.

Coding Index
49.3
#36 of 40 scored
Terminal-Bench
53.9
#36 of 40 scored
Intelligence Index
38.3
#38 of 46 scored
Long-context reasoning
71
reasoning across a long context
SciCode
39.9
scientific coding
GPQA Diamond
86.7
graduate-level QA

usage calculator

How far a month of credits goes on Nemotron 3 Ultra.

Input tokensfresh prompt
800
Output tokensmodel reply
180
Cache read tokensre-read context
50K
cost / request $0.0069 · in $0.60 · out $2.40 · cache $0.12 per M
fresh input 7%output 6%cache reads 87%
Requests / 30 days
1.4K
$10 credits ÷ $0.0069 per request
~289 quick fixes~58 bug fixes~10 feature PRs

what real work costs

Real coding tasks priced end to end on Nemotron 3 Ultra, from a quick lookup to a full-repo agent run.

One agent task
$0.07
180K in at 75% cache hit, 12K out
What you pay
$0.38 /M
all-in across every token that task touched
Sticker input
$0.60 /M
cache reads bill at $0.12 /M instead
TaskTokens in · outNemotron 3 UltraMuse Spark 1.2 ContributorClaude Haiku 4.5
Quick lookup / one-liner8K · 1K$0.0037$0.0003$0.0065
Review a 500-line PR60K · 4K$0.03$0.0027$0.04
Fix a bug (agent loop)180K · 12K$0.07$0.0072$0.12
Refactor a module320K · 20K$0.13$0.01$0.21
Full-repo agent run900K · 45K$0.31$0.03$0.49

frequently asked

What is Nemotron 3 Ultra best for?
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 supports text input and output with a context window of up to 1M tokens. It is suited for long-running agentic workflows, including agent orchestration, coding agents, deep research, and complex enterprise tasks. It is particularly strong at multi-step reasoning and planning, with high-throughput inference designed for high-volume agent pipelines. It is part of the NVIDIA Nemotron family of open models for agentic AI.
How much does Nemotron 3 Ultra cost?
$0.60/M input and $2.40/M output, cache reads $0.12/M. In an agent loop most input is cache-read, so the effective input rate is about $0.26/M.
Which plan do I need?
Available on Go and above.
How do I switch to it?
Run cmd --model nvidia/nemotron-3-ultra-550b-a55b, or type /model in a session and pick it. You can switch mid-session without losing context.

Ship code that matches your taste

Command Code is the AI coding agent that continuously learns your taste. Start for $1.

Benchmarks from Artificial Analysis (v4.1)commandcode.ai/models/nemotron-3-ultra-550b-a55b