Performance Update for Qwen3.8 Flash Next on DGX Spark

X postMiaAnnouncement

Mia (@MiaAI_lab) announced a major performance update for Qwen3.8 Flash Next running on a single DGX Spark, achieving up to 108 tok/s across concurrent streams and extending the KV cache to 1M.

Key points

  1. 46 tok/s for prose generation in a single stream on a single DGX Spark.
  2. 108 tok/s for prose across four concurrent streams.
  3. Prefill speed between 2,000 and 2,200 tok/s across all context lengths.
  4. KV cache size increased to 1M.
  5. Work is underway on a tensor parallelism recipe for tp=2.

Why it matters

Optimising inference throughput and prefill rates on single-node hardware significantly lowers deployment costs and latency for large-context open-weights models.

Analysis

The update highlights strong hardware saturation on a single DGX Spark, scaling output from 46 tok/s in single-stream mode to 108 tok/s under concurrent load. The author noted that multi-GPU tensor parallelism (tp=2) requires a distinct configuration currently under development, and confirmed that dflash2 is not yet supported for Qwen3.8 Flash.

Original post

Mia

@MiaAI_lab · 5 September 2026

MASSIVE performance update to Qwen3.8 Flash Next on a single DGX Spark 🔥 - 46 tok/s for prose, single stream - 108 tok/s for prose across 4 concurrent streams - 2,000–2,200 tok/s prefill across all context sizes - Full numbers in the post below KV cache is now 1M instead of

632 likes67 replies520 bookmarks75.7k views

View on X
More posts in the thread (9)
  1. Full prose tok/s and prefill numbers t.co/hzFWiOxuut
  2. @JDofLondon not for 1 spark, not worth it.
  3. @volatilemarkts @jvr0x Thank you my friend
  4. @Euclides_Netok it's coming !!!!
  5. @CoinUpBTC working on it :)
  6. @jmklex completely diff recipe, now working on tp=2
  7. @ncank There is no dflash2 yet for qwen3.8 flash
  8. @alexellisuk probably old data. 28 tok/s needs to be removed
  9. @wtogami No but I can probably make one. Need?
Selected replies (5)
  • @jmklex @MiaAI_lab What speeds do you get with tp-2 on this recipe?
  • @MichaelGannotti @MiaAI_lab This build is awesome! Set up and had my agents test this morning t.co/zMG80mQYvf
  • @NormanHead94546 @MiaAI_lab Now do tp2
  • @volatilemarkts @MiaAI_lab @jvr0x Great work.
  • @Euclides_Netok @MiaAI_lab NICE! Waiting support for 5060ti

Mentioned links

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • X post:Mike Gannotti

    Qwen3.8-Flash-Next on a Single NVIDIA DGX Spark Outscores Dual-Spark Setup in Benchmark

    A benchmark report from SMF Clearinghouse shows MiaAI's Qwen3.8-Flash-Next (NVFP4) scoring 87.3% on smf-bench Official A running on a single NVIDIA DGX Spark (GB10). The single-node serve outperformed the previous dual-Spark DeepSeek V4 Vision-Exp setup in overall accuracy, achieving a perfect score in coding.

    17bookmarks2.3kviews

    AI & LLMs· News

  • Article:Nemo, DGX Spark & Local Inference Engineer

    Qwen3.8-Flash-Next Scores 87.3% on a Single DGX Spark

    A local benchmark test running Qwen3.8-Flash-Next NVFP4 on a single NVIDIA DGX Spark (GB10) achieved 87.3% on smf-bench Official A, outperforming a previous dual-Spark DeepSeek V4 setup in quality.

    AI & LLMs· Research

  • X post:Mia

    Running DeepSeek v4 Flash on a Single NVIDIA DGX Spark

    MiaAI-Lab published an open-source launcher allowing DeepSeek V4 Flash 0731 (EXL3) to run on a single NVIDIA DGX Spark node instead of requiring a multi-node setup.

    71bookmarks9.8kviews

    AI & LLMs· Tool

  • Repository:MiaAI-Lab/DeepSeek-v4-Flash-One-DGX-Spark

    Running DeepSeek-v4-Flash on a Single DGX Spark

    An open-source launcher and Docker recipe for running DeepSeek V4 Flash 0731 (EXL3) on a single NVIDIA DGX Spark with 128 GiB unified memory.

    328starsPython

    AI & LLMs· Tool

  • Repository:MiaAI-Lab/GLM-5.3-Flash-EXL3-2x-DGX-Sparks

    Serving GLM-5.3 Flash EXL3 on Dual NVIDIA DGX Spark Kits

    Mia's AI Lab provides an optimised vLLM deployment overlay for GLM-5.3-Flash using 4 bpw EXL3 quantisation tailored for 2x NVIDIA GB10 systems.

    431starsPython

    AI & LLMs· Tool

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.