Nemotron 3 Super 120B A12B
View websiteReleased March 2026 under the NVIDIA Nemotron Open Model License, a 120B LatentMoE hybrid activating 12B parameters per token, combining Mamba-2, MoE, and attention layers with Multi-Token Prediction for faster generation. Trained with NVFP4 quantization, covers seven languages, and needs at least 8x H100 80GB to serve.
At a glance
- Context window
- 1M tokens
- Knowledge cutoff
- Jun 2025
- Modalities
- Text Text → Text Text
Capabilities
Function calling
Connect to external tools, APIs, and systems.
Structured output
Return responses in structured formats like JSON.
Pricing by provider
| Provider | Input / 1M tokens | Output / 1M tokens | |
|---|---|---|---|
Geodd
|
$0.09 | $0.50 | View |
Heads up: We do our best to keep these specs & prices accurate. However, cloud costs may fluctuate based on region, usage, and other factors not listed here. These are estimates based on common setups and are for informational purposes only. Always verify current rates & exact specs with the provider before provisioning.
Compare with other models
Estimated prices shown. Actual costs may vary based on context length, batch size, caching, and provider-specific pricing tiers.