---
type: "Article"
title: "MiMo-V2.6-Flash 311B FP8 with vLLM on 4× DGX Spark"
description: "MiMo-V2.6-Flash 311B FP8 with vLLM on 4× DGX Spark: recipe by tonyd2wild with speed, condition and source for every number, weights and license."
resource: "https://www.contextstudios.ai/local-ai/tonyd2wild--mimo-v2.6-flash-dgx-spark-recipe-4x"
language: "en"
generated:
  by: "process:contextstudios-md/1"
  at: "2026-11-11T03:30:07.352Z"
status: "stable"
---

# MiMo-V2.6-Flash 311B FP8 with vLLM on 4× DGX Spark

MiMo-V2.6-Flash 311B FP8 with vLLM on 4× DGX Spark: 59.8 tok/s according to github.com (dataset as of Nov 3, 2026).

- Creator: [tonyd2wild](https://www.contextstudios.ai/local-ai/creators/tonyd2wild.md)
- Engine: vLLM
- Quantization: Original-Checkpoint (FP8), FP8 KV
- Model family: MiMo-V2.6-Flash
- Hardware: dgx-spark x4
- Context: 500000

## All numbers

- 59.8 tok/s (Mixed; Mixed, 1 request, mixed prompt set, with DFlash (per recipe)) — [source](https://github.com/tonyd2wild/MiMo-V2.6-Flash-DGX-Spark-Recipe/blob/main/results/tp4/bench-tp4-gmu85-c1clean.md)
- decode_tps: 59.82 (1 request) — [source](https://github.com/tonyd2wild/MiMo-V2.6-Flash-DGX-Spark-Recipe/blob/main/results/tp4/bench-tp4-gmu85-c1clean.md)
- decode_tps: 136.16 (1 request) — [source](https://github.com/tonyd2wild/MiMo-V2.6-Flash-DGX-Spark-Recipe/blob/main/results/tp4/bench-tp4-gmu85-c1clean.md)
- prefill_tps: 2952.1 (1 request) — [source](https://github.com/tonyd2wild/MiMo-V2.6-Flash-DGX-Spark-Recipe/blob/main/results/tp4/bench-tp4-gmu85-base.md)

## Weights

- [XiaomiMiMo/MiMo-V2.6-Flash-RL](https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL) (main, mit)

[https://github.com/tonyd2wild/MiMo-V2.6-Flash-DGX-Spark-Recipe](https://github.com/tonyd2wild/MiMo-V2.6-Flash-DGX-Spark-Recipe)

## Related

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