Pinned Loading

1. deepseek-v41-tensorfold-spark deepseek-v41-tensorfold-spark Public

   DeepSeek-V4.1-Flash on 2x NVIDIA DGX Spark with the TensorFold engine: 1.8-1.9x code, 2.5x multi-stream, 1.7-1.95x prefill vs the vLLM kit, exact speculative decoding. Work in progress.

   Zig 143 20

2. localrouter localrouter Public

   Fast & opinionated local AI routing: one MCP server that loads local image and video models on demand on your own GPU box (DGX Spark first)

   Zig 102 6

3. glm53-tensorfold-spark glm53-tensorfold-spark Public

   GLM-5.3-Flash (abliterated EXL3) on 2x NVIDIA DGX Spark with the TensorFold engine: 1.8x faster decode than vLLM, 4x256k concurrent threads, byte-exact speculative decoding. Work in progress.

   Python 123 11

4. qwen4-tensorfold-spark qwen4-tensorfold-spark Public

   Recipe (profiles + docs) to serve Qwen3.8-Flash-Next on 1 or 2 DGX Sparks with the Zig TensorFold engine + Compute Lanes. Engine: github.com/jayleaton/TensorFold (branch flashnext-tp1-lanes)

   Dockerfile 17 1

5. go-fast-stack go-fast-stack Public

   Go + Datastar monorepo template (successor to ai-patterns-monorepo)

   Go 11 2

6. ada ada Public

   Forked from pingdotgg/t3code

   TypeScript 20 3