DeepSeek-V4-Flash (official FP8) running across 2x DGX Spark — TP=2, MTP, 200K ctx, recipe + numbers

bjk110 · June 7, 2026, 12:02am

README.md

main

# vLLM Spark — Unified Serving for DGX Spark (GB10)

Unified vLLM serving configuration for NVIDIA DGX Spark (GB10), supporting two
topologies from the same repo / Dockerfile / compose file:

- **Single Spark** (default, zero RDMA setup) — one GB10 box, TP=1.
- **Dual Spark + 200 Gbps RoCE/IB** — two GB10 boxes, Ray, TP=2.

Pick the topology by setting `CLUSTER_MODE=single` (default) or
`CLUSTER_MODE=dual-rdma` in your `.env`. See [`Quick Start`](#quick-start) below.

For release-by-release detail and patch-by-patch status, see
[`CHANGELOG.md`](CHANGELOG.md) and [`PATCH_STATUS.md`](PATCH_STATUS.md).

For a high-level overview of the current repository state and deferred cleanup roadmap,
see [`docs/repository-status.md`](docs/repository-status.md).

## Hardware

| Topology | Node | Role | GPU | Memory | Interconnect |

This file has been truncated. show original

This is the configuration that worked

Topic		Replies	Views
Deepseek v4 Flash on 2 Nodes DGX Spark / GB10 Projects deepseek	71	6574	June 15, 2026
DeepSeek v4 Flash (Aiden Recipe from Reddit) - 1M token session operational, Cuda 12.1 tailored for DGX Spark GB10 DGX Spark / GB10 deepseek	278	13481	July 3, 2026
DeepSeek-V4-Flash on 4× DGX Spark via vLLM (jasl fork, TP=4, RDMA, MTP) — 49–54 tok/s single-stream, full recipe + the traps DGX Spark / GB10 Projects deepseek	3	544	June 19, 2026
Deepseek V4 released DGX Spark / GB10 deepseek	143	16886	May 18, 2026
DeepSeek V4 Flash (1,048,576 Context) on 2x DGX Spark – Custom Sparkrun Recipe DGX Spark / GB10 jetson , deepseek	11	908	June 14, 2026
Fully custom CUDA-native Deepseek 4 Flash optimized for 1x Spark! antirez/ds4 DGX Spark / GB10 Projects gaming , llama , deepseek	77	7959	June 28, 2026
DeepSeekV4-Flash hybrid quant, 1x DGX Spark: antirez's optimized 128 GB MLX recipe ported to vLLM for GB10 DGX Spark / GB10 Projects deepseek	18	2047	May 11, 2026
DeepSeek V4 Flash: Bringing Frontier AI to the Home DGX Spark / GB10 deepseek	11	3660	May 17, 2026
Anyone having luck with Deepseek V4 Flash on Dual Sparks? DGX Spark / GB10 deepseek	13	1425	June 4, 2026
DeepSeek v4 Flash (IQ2XXS) on a single GB10! DGX Spark / GB10 Projects llm , llama , deepseek	13	4159	July 2, 2026

DeepSeek-V4-Flash (official FP8) running across 2x DGX Spark — TP=2, MTP, 200K ctx, recipe + numbers

Related topics