pico-dgx-ds4 social preview
inference-server150MIT

pico-dgx-ds4

DeepSeek V4.1 Flash on 4x DGX Spark: 90/144 tok/s chat/code single-user, 376 tok/s at 16 distinct users, 1M context, images. Fork of blaesus/dsgx10.

Updated Nov 21, 2026
Platforms
Pricing
free-open-source
Status
active
License
MIT

What it does

Core capabilities at a glance

  • Cuda
  • Deepseek
  • DGX Spark
  • Inference Engine
  • LLM Inference
  • Local LLM
  • Multi Node
  • Openai API

Deep dive

The full breakdown - performance, comparisons, and setup

pico-dgx-ds4Link to this section

pico-dgx-ds4 is a local inference server - DeepSeek V4.1 Flash on 4x DGX Spark: 90/144 tok/s chat/code single-user, 376 tok/s at 16 distinct users, 1M context, images. Fork of blaesus/dsgx10.

OverviewLink to this section

Per-commit details and measured before/after numbers are in CHANGES.md; the design is in docs/wide64.md.

Upstream's collectives were built for a cabled ring: each box sends to its two neighbours and they relay to the box opposite. On a switch that second hop is not needed. With 'DSGB10_RING_DIRECT=2' every box writes to all three others itself; buffers, flags and the order of every sum stay the same, so the results are the same bits. A 10 KB all-reduce went 14.5 -> 9.9 us, the engine's step 0.2-0.6 ms shorter: chat +1.3 %, code +1.8 %, math +1.8 % on the 320 held-out prompts, same texts; 16 users with different prompts 364 -> 373 in the same window (live medians 368 -> 376). Two other levers were measured and not built: losslessly coded FP8 weights decoded inside the GEMV (about par with today's kernels, Huffman slower) and a 4-bit drafter body (the lost drafts cancel the shorter pass). The 100 tok/s chat target was not reached. Details: docs/stack5.md.

Three small changes for one user, stacked: chat +3.8 %, code +2.1 %, math +2.5 % on 320 held-out prompts, with the same texts. Details, proofs and what was measured but not built: docs/stack4.md.

pico-dgx-ds4 is open-source, written primarily in Cuda, with 150 GitHub stars under the MIT license. It was last updated on 2026-11-11.

Key capabilitiesLink to this section

From the project's documentation:

  • Each dsgb10 figure is the mean of its sessions on 2026-10-13: six of bench.py (three before the kit's
  • The SGLang kit was measured on this ring on 2026-10-13 from 11:13 to 11:32, in one start: a warm-up session,
  • The MiaAI kit's figures are from its own documents, not measured here. A stack that sums in
  • Against the variant, a step costs the same on the official checkpoint (prefill, the agentic turns and decode
  • The write-up, with every session's figures, the kit's conditions, the correctness checks and how the runs
  • README.md: where it stands, the status of every part, the layout, how to build and run;

How it fits a local-AI stackLink to this section

pico-dgx-ds4 runs on your own hardware, so pair it with a model and a GPU sized to your needs. Use the VRAM calculator to pick a model that fits your card, and see what you can run for hardware guidance. Related local inference servers in the directory:

SourcesLink to this section

Stats from GitHub, 2026-11-14.

Frequently asked

Quick answers to common questions

What is pico-dgx-ds4?

pico-dgx-ds4 is a inference-server tool for local AI workloads. DeepSeek V4.1 Flash on 4x DGX Spark: 90/144 tok/s chat/code single-user, 376 tok/s at 16 distinct users, 1M context, images. Fork of blaesus/dsgx10.

Is pico-dgx-ds4 free and open source?

Yes, pico-dgx-ds4 has 150 GitHub stars and is licensed under MIT. You can self-host it for free on .

What hardware do I need for pico-dgx-ds4?

The hardware requirements depend on which models you run. Check our hardware directory for compatible GPUs and systems. pico-dgx-ds4 has 150 GitHub stars and an active community.

Does pico-dgx-ds4 support GPU acceleration?

pico-dgx-ds4's GPU support depends on your specific setup. Check the documentation for details. For the best performance, pair it with an NVIDIA RTX 4090 or 5090.

What are the best alternatives to pico-dgx-ds4?

Popular alternatives include other inference-server tools in our directory. Browse our full collection at /tool for comparisons, community reviews, and benchmark data to find the right fit for your workflow.

How much does pico-dgx-ds4 cost?

pico-dgx-ds4 is free-open-source. It is completely free and open source to self-host.

Pairs well with

Complementary tools, models, and hardware