---
type: "Article"
title: "Qwen3.8 27.3B with llama.cpp on 1× RTX 3090"
description: "Qwen3.8 27.3B with llama.cpp on 1× RTX 3090: recipe by alesha-pro with speed, condition and source for every number, weights and license."
resource: "https://www.contextstudios.ai/local-ai/alesha-pro--qwen38-27b-mirai-s-mtp3-16gb-1x3090"
language: "en"
generated:
  by: "process:contextstudios-md/1"
  at: "2026-10-23T07:44:37.527Z"
status: "stable"
---

# Qwen3.8 27.3B with llama.cpp on 1× RTX 3090

Qwen3.8 27.3B with llama.cpp on 1× RTX 3090: 57 tok/s according to github.com (dataset as of Sep 29, 2026).

- Creator: [alesha-pro](https://www.contextstudios.ai/local-ai/creators/alesha-pro.md)
- Engine: llama.cpp
- Quantization: Mirai-S-Trellis 2.4-bit + q8_0 KV
- Model family: Qwen3.8
- Hardware: rtx-3090 x1
- Context: 131072

## All numbers

- 57 tok/s (Everyday; Everyday, 1 request, prose prompt, with MTP) — [source](https://github.com/alesha-pro/llama.cpp-mirai-s/blob/master/tools/mirai-s/README.md)
- decode_tps: 57 (1 request) — [source](https://github.com/alesha-pro/llama.cpp-mirai-s/blob/master/tools/mirai-s/README.md)
- decode_tps: 85 (1 request) — [source](https://github.com/alesha-pro/llama.cpp-mirai-s/blob/master/tools/mirai-s/README.md)
- prefill_tps: 806 (1 request) — [source](https://github.com/alesha-pro/llama.cpp-mirai-s/blob/master/tools/mirai-s/README.md)

## Weights

- [alesha-pro/Qwen3.8-27B-S-mirai-GGUF](https://huggingface.co/alesha-pro/Qwen3.8-27B-S-mirai-GGUF) (main, apache-2.0)
- [trymirai/Qwen3.8-27B-S-experimental](https://huggingface.co/trymirai/Qwen3.8-27B-S-experimental) (base, apache-2.0)

[https://github.com/alesha-pro/llama.cpp-mirai-s](https://github.com/alesha-pro/llama.cpp-mirai-s)

## Related

- [Local AI on your hardware.](https://www.contextstudios.ai/local-ai.md)