Nemotron H 47B Reasoning 128K — Hardware Requirements & GPU Compatibility
ChatReasoning
Nemotron H 47B Reasoning 128K is a 46.8B-parameter open language model from NVIDIA in the Nemotron family. At BF16 it needs about 102.94 GB of VRAM — see which GPUs and Macs can run it below.
Nemotron H 47B Reasoning 128K runs on a 128 GB+ Mac or a datacenter GPU
Best local pick → Mac Pro M2 Ultra (128 GB) · BF16 · 107.2 GB · ~5 tok/s
Macs from 128 GB (BF16 · ~5 tok/s)
Other
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Nemotron H 47B Reasoning 128K
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve nvidia/Nemotron-H-47B-Reasoning-128K --max-model-len 16384 --tensor-parallel-size 8vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on H200 (141 GB) (33 GB headroom) · BF16
- Generation speed
- ~30 tok/s
- generation speed
- Cost per 1M output tokens
- $32.91
- per 1M output tokens
How Much VRAM Does Nemotron H 47B Reasoning 128K Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 102.9 GB | 112 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on Instinct MI250X 128 GB |
Selected BF16
Brain floating point 16107.2GBat 16K context
- Weights
- 102.0 GB
- KV cache
- 4.8 GB
- Overhead
- 0.3 GB
Server-class: ~107.2 GB at 16K context · one Instinct MI250X 128 GB
Which Hardware Can Run Nemotron H 47B Reasoning 128K?
BF16 · 102.9 GBNemotron H 47B Reasoning 128K (BF16) requires 102.9 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 134+ GB is recommended. No consumer GPU has enough memory. 18 devices with unified memory can run Nemotron H 47B Reasoning 128K, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Pro M2 Ultra (128 GB).
128–144 GB
fits · 20.9 GB free · up to 32K ctx
4 GPUs · 2 Mac classes · 2 device families
GH200 Grace Hopper Superchip~31Simulate output on GH200 Grace Hopper SuperchipH200 NVL~30Simulate output on H200 NVLH200 SXM~30Simulate output on H200 SXMM2 Ultra*128 GB~5Simulate output on Mac Pro M2 Ultra (128 GB)M5 Max128 GB~4Simulate output on Mac Studio M5 Max (128 GB)GB10~2Simulate output on ASUS Ascent GX10Jetson T5000 (Thor)~2Simulate output on NVIDIA Jetson AGX Thor Developer Kit192+ GB
fits · 84.8 GB free · up to 32K ctx
6 GPUs · 1 Mac class · 2 device families
B200~51Simulate output on B200B300~51Simulate output on B300Instinct MI350X~47Simulate output on Instinct MI350XM3 Ultra*256–512 GB~6Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~169Simulate output on NVIDIA DGX H1008x A100 SXM4~103Simulate output on NVIDIA DGX A100 640GB
By VRAM tier · typical GPUs and Macs at BF16
Consumer GPUs (up to 32 GB) and Macs up to 96 GB: too small for this model (needs ≥ 107.2 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Alternatives to Nemotron H 47B Reasoning 128K
Nemotron family · 42 models →Similar models
3| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Mixtral 8x7B Instruct v0.1Mistral AI | 65.1 | 23.5 GB | ~92 | same size |
| Kimi Linear 48B A3B BaseMoonshot AI | — | 23.4 GB | ~275 | same size |
| Llama 3 3 Nemotron Super 49B V1 5NVIDIA | — | 22.8 GB | ~35 | same size |
Specifications
- Publisher
- NVIDIA
- Family
- Nemotron
- Parameters
- 46.8B
- Release Date
- 2025-05-22
- License
- Other
Frequently Asked Questions
How much VRAM does Nemotron H 47B Reasoning 128K need?
Nemotron H 47B Reasoning 128K requires 102.9 GB of VRAM at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 46.8B × 16 bits ÷ 8 = 93.6 GB
KV Cache + Overhead ≈ 9.3 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run Nemotron H 47B Reasoning 128K?
No — Nemotron H 47B Reasoning 128K requires at least 102.9 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
Can I run Nemotron H 47B Reasoning 128K on a Mac?
Yes — Mac Pro M2 Ultra (128 GB) and 14 other Macs can run Nemotron H 47B Reasoning 128K. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 102.9 GB of usable unified memory (RAM minus macOS overhead).
Can I run Nemotron H 47B Reasoning 128K locally?
Yes — Nemotron H 47B Reasoning 128K can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Nemotron H 47B Reasoning 128K?
At BF16, Nemotron H 47B Reasoning 128K can reach ~47 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 102.9 × 0.65 = ~51 tok/s
Estimated speed at BF16 (102.9 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Nemotron H 47B Reasoning 128K?
At BF16, the download is about 93.58 GB.
Which GPUs can run Nemotron H 47B Reasoning 128K?
No single consumer GPU has enough VRAM to run Nemotron H 47B Reasoning 128K at BF16 (102.9 GB). Multi-GPU or professional hardware is required.
Which devices can run Nemotron H 47B Reasoning 128K?
18 devices with unified memory can run Nemotron H 47B Reasoning 128K at BF16 (102.9 GB), including ASUS Ascent GX10, Mac Pro M2 Ultra (128 GB), Mac Pro M2 Ultra (192 GB), Mac Studio (M3 Ultra, 256GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.