andreaborio/Qwen3.6-35B-A3B-Hebrus-GGUF
Text Generation • Updated • 1.21k • 1
Hebrus ExpertMajor v2 models for Apple Silicon
Exact Qwen, DeepSeek, and GLM ExpertMajor v2 artifacts for Hebrus on Apple Silicon, with scoped RAM, mode, revision, and context contracts.
Note Recommended entry point. Stable Affine4 uses guarded SSD from 16 GiB and is the qwen-v2 default. The same repo also contains an opt-in Q2_K_XL Beta with a 64 GiB floor, qualified only through 32K.
andreaborio/DeepSeek-V4-Flash-Hebrus-GGUF
Text Generation • Updated • 1.89k • 2Note DeepSeek V4 Flash for Apple Silicon with a 64 GiB minimum. Hebrus AUTO may select resident Metal or bounded SSD streaming. Use the exact deepseek-v2 pinned artifact; the separate current 0731 object is opt-in.
andreaborio/GLM-5.2-Hebrus-GGUF
Text Generation • Updated • 392 • 1Note GLM-5.2 ExpertMajor v2 for Apple Silicon with a 64 GiB minimum. Normal AUTO is SSD-only and explicit resident mode is rejected. Start at 8K; the recorded qualified frontier is 32K.