Qwen/Qwen3.6-27B

Image-Text-to-Text • 28B • Updated Apr 24 • 6.73M • • 2.2k

Note Canonical upstream architecture base cited by Prism ML.

prism-ml/Bonsai-27B-gguf

Text Generation • 27B • Updated 22 days ago • 2.67M • 748

Note Primary Bonsai 1-bit GGUF source. The BF16 archival conversion used Bonsai-27B-F16.gguf at revision 0cf7e3d.

prism-ml/Bonsai-27B-mlx-1bit

Text Generation • 2B • Updated 24 days ago • 738k • 205

Note Prism ML native Apple Silicon 1-bit release; upstream reference, not produced by this experiment.

prism-ml/Bonsai-27B-unpacked

27B • Updated 24 days ago • 2.62k • 15

Note Upstream unpacked FP16 safetensors reference for framework compatibility.

prism-ml/Ternary-Bonsai-27B-gguf

Text Generation • 27B • Updated 21 days ago • 796k • • 1.18k

Note Prism ML ternary GGUF comparison point.

prism-ml/Ternary-Bonsai-27B-mlx-2bit

Text Generation • 3B • Updated 24 days ago • 737k • 168

Note Prism ML native Apple Silicon ternary release.

dusterbloom/Bonsai-27B-dSpark-MLX-4bit

Text Generation • 0.3B • Updated 24 days ago • 1.95k • 1

Note MLX 4-bit dSpark speculative-decoding sidecar converted from Prism ML Q4_1 source; separate from this baseline.

TiGa-RCE/Bonsai-27B-MLX-BF16-Config-Repaired

Text Generation • 27B • Updated 12 days ago • 692

Note Mechanical MLX BF16 archival baseline from Prism ML F16 GGUF. Config-only MTP repair documented; source hashes retained.

TiGa-RCE/Bonsai-27B-oQ2e-S32-Smoke

Text Generation • 3B • Updated 12 days ago • 753

Note Bounded 2-bit oQe calibration smoke artifact. Pipeline evidence only, not a quality or production release.

TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke

Text Generation • 4B • Updated 12 days ago • 535

Note Bounded 4-bit oQe calibration smoke artifact. Pipeline evidence only, not a quality or production release.