---
type: "Article"
title: "DeepSeek-V4.1-Flash-1017 FP8/MXFP4 with TensorFold on 4× DGX Spark (switchless ring)"
description: "DeepSeek-V4.1-Flash-1017 FP8/MXFP4 with TensorFold on 4× DGX Spark (switchless ring): recipe by MiaAI-Lab with speed, condition and source for every number, weights and license."
resource: "https://www.contextstudios.ai/local-ai/miaai-lab--deepseek-v4.1-flash-1017-tensorfold-4x-dgx-sparks-switchless"
language: "en"
generated:
  by: "process:contextstudios-md/1"
  at: "2026-11-11T03:30:07.282Z"
status: "stable"
---

# DeepSeek-V4.1-Flash-1017 FP8/MXFP4 with TensorFold on 4× DGX Spark (switchless ring)

DeepSeek-V4.1-Flash-1017 FP8/MXFP4 with TensorFold on 4× DGX Spark (switchless ring): 65.8 tok/s according to github.com (dataset as of Nov 3, 2026).

- Creator: [MiaAI-Lab](https://www.contextstudios.ai/local-ai/creators/miaai-lab.md)
- Engine: TensorFold native (Zig 0.17.0), BertholomusAI dsv41 family (bertholomus/TensorFold deepseek-v41-zig-tp4 @5940208, kit 6033ea4) + recipe zig/patches, TP4 over a switchless ConnectX-7 ring (./start-switchless.sh), DSpark drafts, CUDA graphs
- Quantization: FP8 / MXFP4 (official checkpoint as published, no re-quantization)
- Model family: DeepSeek-V4.1-Flash-1017
- Hardware: dgx-spark x4
- Context: 1048576

## All numbers

- 65.8 tok/s (Everyday; Everyday, 1 request, prose prompt (T = 0.7, thinking on), with DSpark (per recipe)) — [source](https://github.com/MiaAI-Lab/DeepSeek-V4.1-Flash-1017-4x-DGX-Sparks-TensorFold#performance)
- decode_tps: 65.8 (1 request) — [source](https://github.com/MiaAI-Lab/DeepSeek-V4.1-Flash-1017-4x-DGX-Sparks-TensorFold#performance)
- decode_tps: 117.4 (1 request) — [source](https://github.com/MiaAI-Lab/DeepSeek-V4.1-Flash-1017-4x-DGX-Sparks-TensorFold#performance)
- prefill_tps: 4127.2 (1 request) — [source](https://github.com/MiaAI-Lab/DeepSeek-V4.1-Flash-1017-4x-DGX-Sparks-TensorFold#performance)
- ttft_ms: 186 (1 request) — [source](https://github.com/MiaAI-Lab/DeepSeek-V4.1-Flash-1017-4x-DGX-Sparks-TensorFold#performance)

## Weights

- [deepseek-ai/DeepSeek-V4.1-Flash-1017](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017) (main, mit)

[https://github.com/MiaAI-Lab/DeepSeek-V4.1-Flash-1017-4x-DGX-Sparks-TensorFold](https://github.com/MiaAI-Lab/DeepSeek-V4.1-Flash-1017-4x-DGX-Sparks-TensorFold)

## Related

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