Local inference recipes for DGX Spark and Mac Studio — DeepSeek-V4-Flash, GLM-5.3-Flash, MiniMax-H3, Qwen3.8, oMLX Dual-ANE, Anemll/vLLM.
Latest: Keys Big Repo Update — 08/27/2026 (Mac GLM-5.3 Dual-ANE Ablit). Prior: 08/23/2026.
These packs are free and built on a lot of other people's open work. If they save you time or GPU-hours, please consider a donation:
→ GoFundMe — Support Community AI with DGX-Spark, GB10, GPU, and Macs
Thank you.
| Pack | Link |
|---|---|
| Spark GLM-5.3 NVFP4 1M | keys-vLLm.0.27.1-GLM-5.3-Flash-NVFP4-NVFP4KV-1M-Context-Abliterated |
| Mac GLM-5.3 Dual-ANE Ablit | keys-Mac-oMLX-0.6.3.2RC-Dual-ANE-GLM-5.3-Flash-Abliterated-oQ4 |
| Mac Qwen Dual-ANE | keys-Mac-oMLX-0.6.3.2RC-Qwen3.8-27B-Dual-ANE-CPU-Abliterated-oQ4e |
| Mac DSV4F Dual-ANE | keys-Mac-oMLX-0.6.3.2RC-DeepSeekV4F-0731-MXFP4-Abliterated-Dual-ANE-CPU |
| Two-Spark DSV4F 1M | keys-Enhanced-Abliterated-DeepSeekv4-Flash-DSpark-0731-1M-Context-on-Two-DGX-Sparks |
| H3 parallel fleet | keys-MiniMax-H3-Nvidia-Sol-Engine-Two-or-More-DGX-Sparks-Parallel-Processing-Enabled |
| H3 heretic single-Spark | keys-heretic-MiniMax-H3-sol-engine-more-speed-upgrades-upscaler-finish-Single-DGX-Spark |
| Sparkticus dual-boot | keys-DGX-Sparkticus-Ultimate-Power-Pack-Unleashed-Dual-DGX-Sparks-Needed |
| Auto Receipts Studio | keys-Auto-Receipts-Studio |
Each recipe README has a Big thanks block. Full names and links live in each repo's CREDITS section (oMLX, Anemll, MiaAI-Lab, Tony / tonyd2wild, MiniMax, NVIDIA SANA, and others).