I have to retract what I posted as the fastest speed of Qwen 3.8 Flash on a single DGX…
This is a dev post classified by Jev as Hosting & infra (a tutorial), kept by the Dev Radar because it carries real work, not commentary.
I have to retract what I posted as the fastest speed of Qwen 3.8 Flash on a single DGX Spark. Cruz shipped a custom ExLlamaV3 fork and a native engine path. I ran the same model on the same Spark: 102.6 max, 80 tok/s on code, 70 tok/s on prose. My previous benchmark and community recipes was 48.6 tok/s. I ran my full grid on it so you don't have to. This might now be the best way to run this model, without breaking the bank on one DGX Spark. THE QUANT EXL3 at 3.05 bits per weight. Trellis coding packs information more efficiently than plain rounding, so 3 EXL3 bits should land near 4.5
Posted by Yume_X (2.1k followers) 1 h ago · 8 likes · 759 views · view the original post on X. Kept by the Dev Radar as Hosting & infra.
More dev work like this
- How does a global bank navigate regulations and rapid AI cycles without vendor lock-in?… — @RedHat
- Level up your skills while shaping open source priorities! — @CloudNativeFdn
- 🤙 DeepSeek V4.1 Flash. > 100 t/s reached. max 178 t/s stream decode at b1. — @qubitium
- Azure AI Engineer Roadmap Explained! — @AiswaryaVenkit1
- 文科生有救了，妈妈再也不怕我的localhost:8888别人访问不了了！ — @mylifcc
- Kubernetes Node Failure & Recovery in action 👇 — @twtayaan
- 一个 3.3 万 Star 的 Rust 项目，刚刚发布 1.0。 — @mylifcc
- boat — @AniC_dev
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 16.1k posts from 4.9k X accounts over the last 21 days, 1.8k tools, 12 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-20 19:24 UTC. Full method.