Nemotron 3 Ultra API
About
Build advanced reasoning agents with Nemotron 3 Ultra, a specialized AI model optimized for complex logic and chain-of-thought tasks. This model delivers cost-effective pricing at $0.75/1M input and $2.75/1M output tokens, native tool calling support, open weights architecture. Access Nemotron 3 Ultra via the CoreWeave API with up to 262K output tokens.
Capabilities
Input Modalities
text
Output Modalities
text
Reasoning
Yes
Structured Output
Yes
Tool Use
Yes
WeightsOpen
Limits
Context Window
262K
Tokens
Input Limit
262K
Tokens
Max Output
262K
Tokens
Pricing
Standard (per 1M tokens)
Input
$0.75
Output
$2.75
Caching (per 1M tokens)
Read
$0.15
Frequently Asked Questions
Nemotron 3 Ultra by CoreWeave costs $0.75 per 1M input tokens and $2.75 per 1M output tokens. Cached reads cost $0.15 per 1M tokens.
Model Details
ID
nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B
Provider
CoreWeaveFamily
nemotron
Release Date
Jun 4, 2026
Knowledge Cutoff
N/A
API Integration
NPM Package
@ai-sdk/openai-compatible
Environment Variables
WANDB_API_KEY
API Base URL
https://api.inference.wandb.ai/v1