Can Arete 27B run on the RTX 4080?
Yes. At 8K context the best fit is Q3_K_M, using about 13.6 GB of the 15.2 GB usable on the RTX 4080. Estimated speed ceiling: ~54 tokens/sec.
Better quality with offload: 4-bit (Q4_K_M)
Runs with offload 4-bit (Q4_K_M) at 8K context: 53 of 64 layers on the RTX 4080 (about 14.6 GB) and the other 11 in system RAM (about 2.8 GB). Needs a PC with at least 32 GB of system RAM. Estimated speed: 6.4–11 tokens/sec (slow but usable), versus a bandwidth ceiling of ~42 tok/s if it fit fully in VRAM.
llama-server -hf bartowski/NousResearch_Arete-27B-GGUF:Q4_K_M -c 8192 -ngl 53
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, lower -ngl by 2-3; if VRAM is left over, raise 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 | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 11.0 GB Fits | 11.2 GB Fits | 12.7 GB Fits | 18.7 GB No | 68 |
| 3-bit (Q3_K_M) • | 13.4 GB Fits | 13.6 GB Fits | 15.1 GB Tight | 21.1 GB No | 54 |
| 4-bit (Q4_K_M) • | 17.1 GB No | 17.4 GB No | 18.9 GB No | 24.9 GB No | — |
| 5-bit (Q5_K_M) • | 20.4 GB No | 20.6 GB No | 22.1 GB No | 28.1 GB No | — |
| 6-bit (Q6_K) • | 23.1 GB No | 23.4 GB No | 24.9 GB No | 30.9 GB No | — |
| 8-bit (Q8_0) • | 27.8 GB No | 28.1 GB No | 29.6 GB No | 35.6 GB No | — |
| 16-bit (F16/BF16) • | 51.2 GB No | 51.5 GB No | 53.0 GB No | 59.0 GB No | — |
• real GGUF file size from bartowski/NousResearch_Arete-27B-GGUF. Speed is a bandwidth ceiling (717 GB/s), not a benchmark. Methodology.
What fits instead
- Cheapest listed GPU that runs Arete 27B: RX 7900 XTX (24 GB, launch MSRP $999) at Q4_K_M, fully in VRAM at 8K context.
- Largest popular model that fits the RTX 4080: gpt-oss 20B (21B, MoE) at Q8_0, about 12.3 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
Arete 27B on every GPU and Mac · Every model that runs on the RTX 4080 · Other models that fit 16 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/Arete-27B), NVIDIA specs. Model data last checked 2026-11-17; refreshed hourly.