Can Nous Hermes 2 Mixtral 8x7B DPO run on the RTX 3090?
Yes. At 8K context the best fit is Q2_K, using about 17.6 GB of the 22.8 GB usable on the RTX 3090. Estimated speed ceiling: ~196 tokens/sec.
Better quality with offload: 4-bit (Q4_K_M)
Runs with offload 4-bit (Q4_K_M) at 8K context: all attention layers plus the experts of 24 of 32 MoE layers on the RTX 3090 (about 21.1 GB), and the experts of the other 8 layers in system RAM (about 6.4 GB). Needs a PC with at least 32 GB of system RAM. Estimated speed: 12–22 tokens/sec (usable for chat), versus a bandwidth ceiling of ~119 tok/s if it fit fully in VRAM.
llama-server -hf TheBloke/Nous-Hermes-2-Mixtral-8x7B-DPO-GGUF:Q4_K_M -c 8192 -ngl 999 --n-cpu-moe 8
Estimate, not a benchmark. The command downloads the GGUF from Hugging Face. The speed range spans dual-channel DDR4-3200 to DDR5-6000 system RAM; your CPU, RAM speed and llama.cpp version move it. If llama.cpp runs out of GPU memory, raise --n-cpu-moe by 1-2; if VRAM is left over, lower it for more speed. Recent llama.cpp builds can also pick the split automatically. How we estimate offload.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|
| 2-bit (Q2_K) • | 17.1 GB Fits | 17.6 GB Fits | 20.6 GB Tight | 196 |
| 3-bit (Q3_K_M) • | 22.0 GB Tight | 22.5 GB Tight | 25.5 GB No | 151 |
| 4-bit (Q4_K_M) • | 27.5 GB No | 28.0 GB No | 31.0 GB No | — |
| 5-bit (Q5_K_M) • | 31.9 GB No | 32.4 GB No | 35.4 GB No | — |
| 6-bit (Q6_K) • | 36.7 GB No | 37.2 GB No | 40.2 GB No | — |
| 8-bit (Q8_0) • | 47.2 GB No | 47.7 GB No | 50.7 GB No | — |
| 16-bit (F16/BF16) | 88.0 GB No | 88.5 GB No | 91.5 GB No | — |
• real GGUF file size from TheBloke/Nous-Hermes-2-Mixtral-8x7B-DPO-GGUF. Speed is a bandwidth ceiling (936 GB/s), not a benchmark. Methodology.
What fits instead
- Cheapest listed GPU that runs Nous Hermes 2 Mixtral 8x7B DPO: 2x RTX 3090 (2× 24 GB, launch MSRP $2,998) at Q6_K, fully in VRAM at 8K context.
- Largest popular model that fits the RTX 3090: GLM 4.7 Flash (31B, MoE) at Q4_K_M, about 18.3 GB at 8K context.
"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.
More
Nous Hermes 2 Mixtral 8x7B DPO on every GPU and Mac · Every model that runs on the RTX 3090 · Other models that fit 24 GB VRAM · Try other settings in the calculator
Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs
Data: Hugging Face Hub (NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO), NVIDIA specs. Model data last checked 2026-11-18; refreshed hourly.