NVIDIA · First MLX build 20 Mar 2026
Nemotron-Cascade-2-30B-A3B
MLX
Text nvidia-open-model-license
A 32B parameter text generation model from NVIDIA, with 9 MLX builds, from 4-bit to bf16.
- Parameters
- 32B From Hugging Face
- Quantisation
- 4-bit Nemotron-Cascade-2-30B-A3B-4bit
- Memory needed
- 22 GB Estimate, unified memory
- Download
- 17.8 GB Weights, estimated
- Downloads
- 613 This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 9 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 22 GB at 4-bit
8 GB
16 GB
24 GB
32 GB
64 GB
128 GB
Fits with headroomTight, close other appsWon’t fit
Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.
Run it
Terminal · mlx-lm$ pip install -U mlx-lm
$ mlx_lm.generate --model mlx-community/Nemotron-Cascade-2-30B-A3B-4bit --prompt "Hello"MLX builds
9 on mlx-community| Nemotron-Cascade-2-30B-A3B-4bit | 4-bit | 17.8 GB | 22 GB | 613 |
|---|---|---|---|---|
| Nemotron-Cascade-2-30B-A3B-nvfp4 | 4-bit | 17.8 GB | 22 GB | 153 |
| Nemotron-Cascade-2-30B-A3B-mxfp4 | MXFP4 | 17.8 GB | 22 GB | 39 |
| Nemotron-Cascade-2-30B-A3B-5bit | 5-bit | 21.7 GB | 27 GB | 40 |
| Nemotron-Cascade-2-30B-A3B-6bit | 6-bit | 25.7 GB | 31 GB | 207 |
| Nemotron-Cascade-2-30B-A3B-mlx-6bit | 6-bit | 25.7 GB | 31 GB | 44 |
| Nemotron-Cascade-2-30B-A3B-8bit | 8-bit | 33.6 GB | 40 GB | 247 |
| Nemotron-Cascade-2-30B-A3B-mxfp8 | MXFP8 | 33.6 GB | 40 GB | 28 |
| Nemotron-Cascade-2-30B-A3B-bf16 | bf16 | 63.2 GB | 73 GB | 37 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.