How fast does Qwen3.8 27B run on the NVIDIA RTX 3080?

Qwen3.8 27B runs at about 45.9 tokens/s on the NVIDIA RTX 3080 (measured, UD-IQ2_XXS, CUDA, llama-bench). Based on 2 measured run(s), each linked to its public source.

Does Qwen3.8 27B UD-IQ2_XXS fit in the NVIDIA RTX 3080's memory?

Estimate: the UD-IQ2_XXS file is 7.3 GB and needs about 8.7 GB with an ~8K context; the NVIDIA RTX 3080 has about 9.5 GB usable, so it fits.

Qwen3.8 27B › NVIDIA RTX 3080

Qwen3.8 27B on the NVIDIA RTX 3080

Qwen3.8 27B runs at about 45.9 tokens/s on the NVIDIA RTX 3080 (measured, UD-IQ2_XXS, CUDA, llama-bench).

Best measured: 45.9 tokens/s generation and 1,132 tokens/s prompt processing (UD-IQ2_XXS, CUDA, FA on).

For comparison, our size-scaling estimate for this combination is 76.7 tok/s.

Every run

Quant	Backend	Generate tok/s	Prompt tok/s	Test	Build / by / notes	Source
UD-IQ2_XXS	CUDA FA on	45.9	1,132	tg128	llama-bench on the RTX 3080 10GB row's file (unsloth UD-IQ2_XXS), headless, b10472.	qwen38-mtp community table (GitHub)
UD-IQ2_XXS	CUDA	45.1	–	decode, llama-server, MTP off	Baseline (spec-off) column of a paired MTP A/B: llama-server single stream (--parallel 1), probe.py median, not llama-bench.	qwen38-mtp community table (GitHub)

tg128 = generating 128 tokens; pp512 = reading a 512-token prompt (llama-bench). LocalScore averages several prompt and output lengths, so its numbers run lower than tg128. "decode, llama-server" = single-stream generation speed measured against a running llama-server (no speculative decoding).

Check prices

Amazon Newegg

As an Amazon Associate I earn from qualifying purchases.

No card yet?

Try it on a rented GPU (RunPod)

Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.

All results for the NVIDIA RTX 3080 · Qwen3.8 27B on other hardware

Embed this result

HTML

<a href="https://rigbench.pages.dev/model/qwen3-8-27b/on/nvidia-rtx-3080/"><img src="https://rigbench.pages.dev/badge/qwen3-8-27b--nvidia-rtx-3080.svg" alt="Qwen3.8 27B · NVIDIA RTX 3080: 45.9 tok/s (UD-IQ2_XXS, measured)"></a>

Markdown (GitHub README, forums)

[![Qwen3.8 27B · NVIDIA RTX 3080: 45.9 tok/s (UD-IQ2_XXS, measured)](https://rigbench.pages.dev/badge/qwen3-8-27b--nvidia-rtx-3080.svg)](https://rigbench.pages.dev/model/qwen3-8-27b/on/nvidia-rtx-3080/)