Mau “Rules without Rulers” Ledford
@krunkosaurus
Decentralized AI, Local AI, co-founder @Sogni_protocol
3.4K Following 5.7K Followers
Qwen3.8-Flash-Next on ONE DGX Spark.
~180B MoE, Thinking MAX, full 262K context, vision. Decode holds 48-62 t/s from 6K all the way to 240K.
If you only have 1 Spark, this is probably the Qwen 3.8 you want over 27B.
Recipe by @MiaAI_lab on TensorFold
Show more
