Spaces:

halley-ai
/

README

No application file

sebastavar commited on Aug 17, 2025

Commit

0333bcf

verified ·

1 Parent(s): 3dd4aec

Update README.md

Files changed (1) hide show

README.md CHANGED Viewed

@@ -17,11 +17,11 @@ High-quality, Apple-Silicon–optimized **MLX** builds, tools, and evals — foc
 ## 🚀 Featured models
-| Repo | Bits / GS | Type | Footprint | Notes |
 |---|---:|---|---:|---|
-| **HalleyAI/gpt-oss-20b-MLX-4bit-gs32** | Q4 / 32 | MLX (Metal) | ~13.1 GB | Best speed on 32 GB; near-baseline quality (+1.81% PPL vs 8-bit) |
-| **HalleyAI/gpt-oss-20b-MLX-6bit-gs32** | Q6 / 32 | MLX (Metal) | ~18.4 GB | Near-Q8 fidelity (-0.51% PPL vs 8-bit) |
-| **Reference (8-bit)** | Q8 / 32 | MLX (Metal) | — | Use upstream: `lmstudio-community/gpt-oss-20b-MLX-8bit` |
 > **Format:** MLX (not GGUF). For Linux/Windows or non-MLX stacks, use a GGUF build with llama.cpp.

 ## 🚀 Featured models
+| Repo | Bits / GS | Footprint | Notes |
 |---|---:|---|---:|---|
+| **HalleyAI/gpt-oss-20b-MLX-4bit-gs32** | Q4 / 32 | ~13.1 GB | Best speed on 32 GB; near-baseline quality (+1.81% PPL vs 8-bit) |
+| **HalleyAI/gpt-oss-20b-MLX-6bit-gs32** | Q6 / 32 | ~18.4 GB | Near-Q8 fidelity (-0.51% PPL vs 8-bit) |
+| **Reference (8-bit)** | Q8 / 32 | — | Use upstream: `lmstudio-community/gpt-oss-20b-MLX-8bit` |
 > **Format:** MLX (not GGUF). For Linux/Windows or non-MLX stacks, use a GGUF build with llama.cpp.