Skip to content

Text model · Nemotron 3

NE Nemotron Cascade 2 30B-A3B: RAM and VRAM requirements

Nemotron Cascade 2 30B-A3B needs about 25.1 GB to run at Q4_K_M (the Q4_K_M GGUF file is ~23.03 GB to download; KV cache and overhead add the rest), or about 33.4 GB at Q8_0. The lightest hardware that runs it is Nvidia GeForce RTX 5090 (32GB).

Nemotron 3 family · 30B params (Mixture-of-Experts: activates only 3B of 30B params per token, so generation is faster than the total size suggests) · released Mar 2026.

Shopping for hardware? See what runs Nemotron Cascade 2 30B-A3B →

License nvidia-open-model-license · Conditional ↓ 48.2K/mo ♥ 522 on HuggingFace
Q4_K_M GGUF
23.03 GB
Q8_0 GGUF
31.28 GB
Memory @ Q4 (4k)
~25.1 GB
Context
256 k

Q4_K_M and Q8_0 are GGUF download (file) sizes. Memory @ Q4 includes the KV cache and overhead, so it runs larger than the file on disk.

Will it run on your device?

Nemotron Cascade 2 30B-A3B runs on 9 of 40 tracked devices at Q4_K_M.

9 run well 0 tight fit 31 too small
Fit check Q4_K_M
No, not enough memory
needs 25.1 GB usable 10.5 GB
See the full breakdown

Memory breakdown

Weights (Q4_K_M)23.03 GB
+
KV cache (4k)1.3 GB
+
Overhead0.8 GB
=
Total25.1 GB

How context length changes it

4k context ~25.1 GB 32k context ~34.3 GB 128k context ~65.9 GB

Longer context grows the KV cache, which for Nemotron Cascade 2 30B-A3B is sized by its 3B active params, not the full 30B. It needs ~25.1 GB at 4k and ~65.9 GB at 128k.

Quantization sizes

GGUF quantson disk
Quantization Size on disk
Q2_K 12.6 GB est
Q3_K_M 14.7 GB est
Q4_K_M (default) 23.03 GB
Q5_K_M 21.4 GB est
Q6_K 24.6 GB est
Q8_0 31.28 GB
FP16 60 GB est

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

Ollama
$ ollama run nemotron-cascade-2:30b
llama.cpp
$ llama-cli -hf bartowski/nvidia_Nemotron-Cascade-2-30B-A3B-GGUF:Q4_K_M
LM Studio
$ lms get bartowski/nvidia_Nemotron-Cascade-2-30B-A3B-GGUF

Which devices can run Nemotron Cascade 2 30B-A3B?

Same job, different size

Pick by what fits your memory: step down to free up VRAM, or step up if you have headroom.

Similar models

Head-to-head

FAQ

How much VRAM or RAM does Nemotron Cascade 2 30B-A3B need?

At Q4_K_M, Nemotron Cascade 2 30B-A3B needs about 25.1 GB (weights ~23.03 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~33.4 GB.

What is the Q4_K_M GGUF file size of Nemotron Cascade 2 30B-A3B?

The Q4_K_M GGUF file is about 23.03 GB to download, and the Q8_0 GGUF is about 31.28 GB. That is the weights file on disk; to run it you also need room for the KV cache and overhead, so budget ~25.1 GB of memory at Q4_K_M.

Can Nemotron Cascade 2 30B-A3B run on a laptop?

Nemotron Cascade 2 30B-A3B is large; you need a high-memory Mac or multi-GPU setup at Q4_K_M.

Is Nemotron Cascade 2 30B-A3B cheaper to run because it is a MoE model?

It is faster, not lighter. Nemotron Cascade 2 30B-A3B activates only 3B of 30B params per token (so it runs quickly), but all experts must stay in memory, so it still needs memory for the full 30B.

Can I use Nemotron Cascade 2 30B-A3B commercially?

Conditionally. NVIDIA Open Model License; check the terms for commercial use.

Understand the numbers

Short guides to the ideas behind Nemotron Cascade 2 30B-A3B's memory and quant figures.

Post-trained (SFT+RL) from the Nemotron-3-Nano-30B-A3B base for reasoning and agentic tasks; NVIDIA reports gold-medal IMO/IOI results. Same 30B MoE architecture as Nemotron-3-Nano-30B-A3B (its own card rounds active params to 3B vs the base's 3.5B) but a distinct checkpoint, not a duplicate listing. Up to 1M context. Sizes from bartowski's GGUF repo. NVIDIA Open Model License.

Sources

Last validated 2026-08-03. Memory figures are estimates. See methodology.