NVIDIA: Nemotron 3 Super
by NVIDIANVIDIA: Nemotron 3 Super is a large language model from NVIDIA. It costs $0.300 per million input tokens and $2.40 per million output tokens. Its context window is 1.0M tokens.
- Input / 1M tokens
- $0.300
- Output / 1M tokens
- $2.40
- Cached input / 1M
- $0.150
- Context window
- 1.0M
On repeated prefixes
tokens
Who serves it cheapest
3 hosts serve Nemotron 3 Super. Same weights, same API - the price difference is pure margin and routing.
| Provider | Input / 1M | Output / 1M | Context | Throughput | Uptime 24h |
|---|---|---|---|---|---|
| DeepInfraCheapestbf16 | $0.085 | $0.400 | 262K | - | 100.0% |
| DekaLLMfp8 | $0.080 | $0.450 | 262K | - | 99.4% |
| Crusoefp8 | $0.300 | $2.40 | 1.0M | - | 100.0% |
The spread between DeepInfra and Crusoe is 5.0× for identical weights. Quantization and context limits differ, so check both columns before switching.
Benchmarks
Independent scores published alongside the catalogue.
- Coding index
- 37.7
- Agentic index
- 1.7
About Nemotron 3 Super
NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications. Built on a hybrid Mamba-Transformer...
Specifications
| Model ID | nvidia/nemotron-3-super-120b-a12b |
|---|---|
| Provider | NVIDIA |
| Context window | 1.0M tokens |
| Max output | 262K tokens |
| Input modalities | text |
| Output modalities | text |
| Knowledge cutoff | - |
| Open weights | Yes - nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8 |
| Released | March 11, 2026 |
Cheaper alternatives
Models that cost less than Nemotron 3 Super while keeping at least half its context window and every input modality it supports.
Frequently asked
How much does NVIDIA: Nemotron 3 Super cost?
$0.300 per million input tokens and $2.40 per million output tokens. Cached input reads cost $0.150 per million tokens.
What is the context window of NVIDIA: Nemotron 3 Super?
1.0M tokens, with up to 262K tokens of output per request.
Which provider serves NVIDIA: Nemotron 3 Super cheapest?
DeepInfra at $0.085 per million input tokens - 5.0× cheaper than Crusoe, the most expensive of the 3 hosts serving it.
Confirm against the source: NVIDIA official pricing.