RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?API pricesVRAM calculatorNew

Can Qwen3.8 27B run on the RTX 4080 Super?

Yes. At 8K context the best fit is Q2_K, using about 11.5 GB of the 15.2 GB usable on the RTX 4080 Super. Estimated speed ceiling: ~68 tokens/sec.

Better quality with offload: 4-bit (Q4_K_M)

Runs with offload 4-bit (Q4_K_M) at 8K context: 52 of 64 layers on the RTX 4080 Super (about 14.5 GB) and the other 12 in system RAM (about 3.1 GB). Needs a PC with at least 32 GB of system RAM. Estimated speed: 5.9–10 tokens/sec (slow but usable), versus a bandwidth ceiling of ~42 tok/s if it fit fully in VRAM.

llama-server -hf bartowski/Qwen3.8-27B-GGUF:Q4_K_M -c 8192 -ngl 52

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.

+ 0.9 GB with image input. The figures on this page are for text only. Sending images also loads the vision projector (mmproj) file: Q2_K at 8K context then needs about 12.3 GB (still fits). Calculator with image input on.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •11.2 GB Fits11.5 GB Fits13.0 GB Fits19.0 GB No68
3-bit (Q3_K_M) •13.6 GB Fits13.9 GB Tight15.4 GB No21.4 GB No55
4-bit (Q4_K_M) •17.4 GB No17.6 GB No19.1 GB No25.1 GB No—
5-bit (Q5_K_M) •20.6 GB No20.9 GB No22.4 GB No28.4 GB No—
6-bit (Q6_K) •23.3 GB No23.6 GB No25.1 GB No31.1 GB No—
8-bit (Q8_0) •28.2 GB No28.5 GB No30.0 GB No36.0 GB No—
16-bit (F16/BF16) •52.0 GB No52.3 GB No53.8 GB No59.8 GB No—

• real GGUF file size from bartowski/Qwen3.8-27B-GGUF. Speed is a bandwidth ceiling (736 GB/s), not a benchmark. Methodology.

What fits instead

"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

Qwen3.8 27B on every GPU and Mac · Every model that runs on the RTX 4080 Super · 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 (Qwen/Qwen3.8-27B), NVIDIA specs. Model data last checked 2026-11-17; refreshed hourly.