Nemotron 70B Instruct — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
A friendly and highly capable model. Great for general purpose chat where nuance and helpfulness are priority.
Nemotron 70B Instruct needs about 43 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
VRAM and speed by quantization
Quoted against NVIDIA RTX 4090 (24 GB). Includes the KV cache at 8K context, so it reads higher than the headline figure.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 26.7 GB | ~4 tok/s | Offload | |
| Q3_K_M 3.41 bpw | 33.6 GB | ~3 tok/s | Offload | |
| Q4_K_M 4.83 bpw | 46.1 GB | ~2 tok/s | Offload | |
| Q5_K_M 5.67 bpw | 53.5 GB | ~2 tok/s | Offload | |
| Q6_K 6.56 bpw | 61.4 GB | — | Too big | |
| Q8_0 8.50 bpw | 78.5 GB | — | Too big | |
| F16 16.00 bpw | 144.7 GB | — | Too big |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Nemotron 70B Instruct VRAM calculator →
Get Nemotron 70B Instruct running
The cheapest catalogued GPU that runs Nemotron 70B Instruct is the NVIDIA DGX Spark (128 GB).
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
How to run Nemotron 70B Instruct
Install Ollama, then run:
ollama run nemotronSpecifications
Curated — A hand-written entry from before this catalogue recorded its sources. The figures are long-standing but their provenance is not on file.
- Parameters
- 70 Billion
- Context window
- 128,000
- Architecture
- Dense
- Provider
- NVIDIA
- Licence
- Llama Community
- Practical deployment
- Workstation (large unified memory or pro GPU)
- Specified at
- Q4_K_M
- System RAM
- 64 GB
- Record updated
- 2024-10-15
Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Quality and use cases
Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.
Can I run Nemotron 70B Instruct on my GPU?
- Nemotron 70B on AMD Radeon RX 7900 XT
- Nemotron 70B on AMD Radeon RX 7900 XTX
- Nemotron 70B on AMD Ryzen AI Max+ 395
- Nemotron 70B on Apple M1 Max
- Nemotron 70B on Apple M1 Pro
- Nemotron 70B on Apple M2
- Nemotron 70B on Apple M2 Max
- Nemotron 70B on Apple M2 Pro
- Nemotron 70B on Apple M3
- Nemotron 70B on Apple M3 Max
- Nemotron 70B on Apple M3 Pro
- Nemotron 70B on Apple M4
- Nemotron 70B on Apple M4 Pro
- Nemotron 70B on Apple M5
- Nemotron 70B on Apple M5 Pro
- Nemotron 70B on NVIDIA A100 80GB (PCIe)
- Nemotron 70B on NVIDIA GeForce RTX 3090
- Nemotron 70B on NVIDIA GeForce RTX 4090
- Nemotron 70B on NVIDIA GeForce RTX 5090
- Nemotron 70B on NVIDIA H100 80GB (PCIe)
- Nemotron 70B on NVIDIA L40S
- Nemotron 70B on NVIDIA RTX 6000 Ada Generation
- Nemotron 70B on NVIDIA RTX PRO 6000 Blackwell
Nemotron 70B Instruct — frequently asked questions
How much VRAM does Nemotron 70B Instruct need?
About 43 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Nemotron 70B Instruct run on an RTX 4090 (24 GB)?
No. Nemotron 70B Instruct needs about 43 GB at Q4_K_M, more than a single RTX 4090's 24 GB. It needs a larger card, several GPUs, or Apple Silicon with enough unified memory — or it runs with part of the weights offloaded to system RAM, which is much slower.
How do I run Nemotron 70B Instruct locally?
Install Ollama and run `ollama run nemotron`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.