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.
- 133
- Cuda
- MIT
- 3 d ago
Picom-code
personal acc
Built from Picom-code's public GitHub profile, Austin. Not on Take My Repo yet, so there are no posts or updates here.
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.
Qwen3.8-Flash-Next on one DGX Spark, official NVIDIA FP4 weights, exact output: 54 chat / 88 code tok/s, 66.9 on 320 held prompts, 288 tok/s at 16 users. TensorFold + MiaAI-Lab recipe + 29 patches.
Agent CLI — hard fork of opencode with cross-device sync and observability.
the lean agent harness — your models, your machine. terminal · desktop · web, built on Hermes