glm-5.3-one-spark
GLM-5.3-Flash EXL3 2.05 on one DGX Spark — TP1 vLLM, DFlash2 K5, concurrency, 262K context, and published fidelity measurements
- 57
- Python
- MIT
- 8 wk ago
gitcommit90
cd /opt/github/repos
Built from gitcommit90's public GitHub profile. Not on Take My Repo yet, so there are no posts or updates here.
GLM-5.3-Flash EXL3 2.05 on one DGX Spark — TP1 vLLM, DFlash2 K5, concurrency, 262K context, and published fidelity measurements
A personal AI workspace on your own Cloudflare account, with persistent memory, channel computers, and your choice of AI provider.
Local AI router for macOS and Linux: accounts, models, named routes, and fallback behind one endpoint.
Measured Laguna S 2.1 NVFP4 + DFlash vLLM package for NVIDIA DGX Spark (GB10)
SenseNova Loop 9B NVFP4 + MTP on DGX Spark: calibrated checkpoint recipe, native FP4 runtime, reproducible evaluation and figures.
Qwen3.8-Flash-Next NVFP4 + dense FP8, MTP K3 and C6 on one NVIDIA DGX Spark
Ripple-65B NVFP4 + DFlash2 on one NVIDIA DGX Spark: tree verification, fused MLP, attention fix, reproducible benchmarks and honest limits.
Local Arena leaderboard explorer with historical rankings and OpenRouter pricing
Qwen3.8-Flash-Next UD-IQ4_XS GGUF on one NVIDIA DGX Spark: 262K context, vision, thinking controls, and measured benchmarks.
vLLM deployment for nvidia/Qwen3.6-35B-A3B-NVFP4 on DGX Spark / GB10 — 10 concurrent sessions, measured benchmarks
vLLM deployment for InternScience/Agents-A1 on DGX Spark / GB10 — 10 concurrent, measured benchmarks
vLLM deployment for nvidia/diffusiongemma-26B-A4B-it-NVFP4 on DGX Spark / GB10 — 10 concurrent, measured benchmarks
llama.cpp deployment for ivanfioravanti DeepSeek-V4-Flash Q4 GGUF on DGX Spark / GB10 — measured single-stream; multi-concurrent still broken
AngelSlim Hy3 IQ1_M MTP ready-to-run package for NVIDIA DGX Spark / GB10 (honest measured numbers)
Terminal-style dashboard for Artificial Analysis model rankings (Intelligence, Coding, Agentic) — live refresh, rank movement, pricing. Zero-dependency Node.