Clover-35B-A3B, a diffusion LLM, decodes at 25.8 tok/s in BF16 on one DGX Spark (128 GB)
This is a dev post classified by AGK Intelligence as AI dev tools (a tool drop), kept by the Dev Radar because it carries real work, not commentary.
Clover-35B-A3B, a diffusion LLM, decodes at 25.8 tok/s in BF16 on one DGX Spark (128 GB) 1 DGX Spark, 128 GB Decode, code: 25.7 tok/s Decode, prose: 26.0 tok/s 500 words (code and prose): about 27.0 s First token at 8K: 1.4 s First token at 32K: 6.6 s First token at 128K: 40.8 s BF16 (16 bpw), SGLang nightly of 18 October plus the authors' open pull request 46927 I changed 7 lines of its device allow-list so it runs on the GB10, the pull request was validated on H200 only 131,072 window, 4 slots (the value in the authors' own test script), the 128K prompt has 130,048 tokens Speed test, medi
Posted by Niklas (2.8k followers) 1 h ago · 5 likes · 364 views · view the original post on X. Kept by the Dev Radar as AI dev tools.
More dev work like this
- The latest llama.cpp release made 2x Sparks work together on the same model ... and… — @TeksEdge
- ChatGPT Images 2.6 sets the scene. Tesseract brings it to life. — @trymirage
- Your coding agent can help script your TikToks. You still shoot the video. — @DanKornas
- 1 prompt and Images 2.6 made this — @vikktorrrre
- Is this a Deepseek reference? 🐋 — @harjjotsinghh
- BREAKING 🔥: OpenAI released GPT‑Image‑2.6 Flare and Sunburst image models on APIs and… — @testingcatalog
- OpenAI drop New GPT-Images 2.6. — @0x0SojalSec
- OPENAI 🚀 — @aitrackerbot
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.6k posts from 5k X accounts over the last 21 days, 9.3k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-05 00:17 UTC. Full method.