Nous Hermes 2 Mixtral 8x7B DPO — Hardware Requirements & GPU Compatibility
Chat
Nous-Hermes-2-Mixtral-8x7B-DPO is Nous Research's instruction-tuned chat model built on top of Mistral AI's Mixtral-8x7B-v0.1 Mixture-of-Experts base, trained through supervised fine-tuning followed by direct preference optimization (DPO) on over a million largely GPT-4-generated examples plus other curated open data. It uses the ChatML prompt format for structured multi-turn dialogue and system prompts, and improved on GPT4All, AGIEval, and BigBench benchmarks over both the base Mixtral model and Mistral's own Mixtral-Instruct. With 8 experts and 2 active per token, roughly 12.9 billion active out of 46.7 billion total, it needs a high-end consumer GPU or multi-GPU setup once quantized. Context length is 32,768 tokens. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use. It was published in January 2024, alongside an SFT-only sibling release for comparison.
Usable on a 24 GB GPU
Runs from 16 GB (IQ2_XXS · 16K)Best on 24 GB (IQ3_M · ~92 tok/s · 16K)Macs from 24 GB (IQ2_M · ~37 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Nous Hermes 2 Mixtral 8x7B DPO
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run nous-hermes2-mixtral:8x7b-dpo-q4_K_MQ4_K_M · needs ~30.5 GB · 16K context
Download: nous-hermes-2-mixtral-8x7b-dpo.Q4_K_M.gguf28.45 GBTheBloke
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 48 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on A100 40GB (9 GB headroom) · Q4_K_M
- Generation speed
- ~105 tok/s
- generation speed
- Cost per 1M output tokens
- $1.62
- per 1M output tokens
How Much VRAM Does Nous Hermes 2 Mixtral 8x7B DPO Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 20.4 GB | 24.5 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 16K context · ~97 tok/s | |
| 21 GB | 25 GB | 3-bit small — 3-bit small quantization | Fits: 16K context · ~94 tok/s | |
| 23.3 GB | 27.4 GB | 3-bit medium — 3-bit medium quantization | Fits: 4K context · ~86 tok/s | |
| 23.9 GB | 28 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 2K context · ~85 tok/s | |
| 28.6 GB | 32.6 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Does not fit: needs 30.5 GB | |
| 33.8 GB | 37.9 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Does not fit: needs 35.7 GB | |
| 39.1 GB | 43.1 GB | 6-bit, very good — 6-bit quantization, very good quality | Does not fit: needs 41 GB | |
| 47.3 GB | 51.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 49.2 GB |
Selected Q4_K_M
4-bit medium quantization30.5GBat 16K context
- Weights
- 28.0 GB
- KV cache
- 2.2 GB
- Overhead
- 0.3 GB
Doesn't fit RTX 4090 (reference) · needs 6.5 GB more, IQ3_M fits
≥50 tok/s needs ≥664 GB/s · RTX 5090 ~117
Or .
Which Hardware Can Run Nous Hermes 2 Mixtral 8x7B DPO?
Q4_K_M · 28.6 GBNous Hermes 2 Mixtral 8x7B DPO (Q4_K_M) requires 28.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 38+ GB is recommended. Using the full 33K context window can add up to 4.0 GB, bringing total usage to 32.6 GB. 1 GPU can run it, including NVIDIA GeForce RTX 5090. 78 devices with unified memory can run Nous Hermes 2 Mixtral 8x7B DPO, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M5 Max (36 GB).
20 GB
Q4_K_M · best fit IQ2_M (18.2 GB)
1 GPU
RX 7900 XT~63Simulate output on RX 7900 XT32–40 GB
fits · 1.5 GB free · up to 16K ctx
1 GPU · 2 Mac classes · 1 device family
RTX 5090~117Simulate output on RTX 5090M5 Max36 GB~32Simulate output on Mac Studio M5 Max (36 GB)M3 Pro*36 GB~12Simulate output on MacBook Pro 14" M3 Pro (36 GB)Orin~16Simulate output on NVIDIA Jetson AGX Orin 32GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
How good is Nous Hermes 2 Mixtral 8x7B DPO?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
This is the single board showing results for this model; badges show who ran the test.
- LMArena Text
LMArena · Oct 2026Arena
Arena1164.5
#165/188 open
Alternatives to Nous Hermes 2 Mixtral 8x7B DPO
Mixtral family · 7 models →Similar models
3| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| 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 |
| Nemotron H 47B Reasoning 128KNVIDIA | — | 107.2 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- Nous Research
- Family
- Mixtral
- Parameters
- 46.7B
- Active parameters
- 12.9B per token
- Architecture
- MixtralForCausalLM
- Context Length
- 32,768 tokens
- Vocabulary Size
- 32,002
- Release Date
- 2024-01-11
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Nous Hermes 2 Mixtral 8x7B DPO need?
Nous Hermes 2 Mixtral 8x7B DPO requires 28.6 GB of VRAM at Q4_K_M, or 94.0 GB at BF16. Full 33K context adds up to 4.0 GB (32.6 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 46.7B × 4.8 bits ÷ 8 = 28 GB
KV Cache + Overhead ≈ 0.6 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.
KV Cache + Overhead ≈ 4.6 GB (at full 33K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run Nous Hermes 2 Mixtral 8x7B DPO?
Yes, at Q4_0 (23.9 GB) or lower. Higher quantizations like Q3_K_L (24.5 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for Nous Hermes 2 Mixtral 8x7B DPO?
For Nous Hermes 2 Mixtral 8x7B DPO, Q4_K_M (28.6 GB) offers the best balance of quality and VRAM usage. Q5_0 (29.8 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 13.4 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Nous Hermes 2 Mixtral 8x7B DPO on a Mac?
Yes — Mac Mini M5 Pro (24 GB) and 106 other Macs can run Nous Hermes 2 Mixtral 8x7B DPO, the smallest at IQ2_M. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 28.6 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio M5 Max (36 GB).
Can I run Nous Hermes 2 Mixtral 8x7B DPO locally?
Yes — Nous Hermes 2 Mixtral 8x7B DPO can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Nous Hermes 2 Mixtral 8x7B DPO?
At Q4_K_M, Nous Hermes 2 Mixtral 8x7B DPO can reach ~124 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 7.9 GB active ÷ (8000 × 0.65) = 1.52 ms, plus 32 layers × 0.055 ms = 1.76 ms, so 1000 ÷ 3.28 ms = ~305 tok/s
Estimated speed at Q4_K_M (28.6 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Nous Hermes 2 Mixtral 8x7B DPO?
At Q4_K_M, the download is about 28.02 GB. The full-precision BF16 version is 93.41 GB. The smallest option (IQ2_XXS) is 12.84 GB.
Which GPUs can run Nous Hermes 2 Mixtral 8x7B DPO?
1 consumer GPU can run Nous Hermes 2 Mixtral 8x7B DPO at Q4_K_M. Top options include NVIDIA GeForce RTX 5090.
Which devices can run Nous Hermes 2 Mixtral 8x7B DPO?
78 devices with unified memory can run Nous Hermes 2 Mixtral 8x7B DPO at Q4_K_M (28.6 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.