Downloads · 30 days
42
9% of all-time downloads
Geer-Qwen3.8-27B-6bit-MLX
ilyakam/Geer-Qwen3.8-27B-6bit-MLX
Geer-Qwen3.8-27B-6bit-MLX is a image-text-to-text model from ilyakam. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for mlx. The card lists the license as apache-2.0.
This is Geer's independently produced, protected 6-bit MLX conversion of Qwen3.8-27B for local agentic coding on Apple Silicon. It preserves the multimodal tower and native MTP head. It is not a new foundation model.
All-time downloads
477
Public
Parameters
27.8B
26.7 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors26.6 GB · 100%
Parameter types
How the weights are stored.
U3224.4B · 88%
Try a prompt
Base models
Model card
From the Hugging Face model README
Geer Qwen 3.8 27B (6-bit MLX)
This is Geer's independently produced, protected 6-bit MLX conversion of Qwen3.8-27B for local agentic coding on Apple Silicon. It preserves the multimodal tower and native MTP head. It is not a new foundation model.
The build uses upstream revision
1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0, affine quantization with group
size 64, and pinned MLX conversion dependencies. The vision tower, embeddings,
output head, norms, and MTP tensors remain unquantized. The included source and
build manifests record exact file hashes.
Upstream benchmark results describe the BF16 checkpoint, not this conversion. Geer uses this artifact on 64 GB and larger Macs with a 128K context ceiling at 64 GB and Qwen's native 256K ceiling at 96/128 GB. Native MTP is calibrated per installation and remains user-configurable. The 48 and 64 GB profiles remain provisional until exercised on real hosts of those sizes.
- Task
- Image-Text-to-Text
- Library
- mlx
- Type
- qwen3_5
- License
- apache-2.0
- Languages
- en
- Created
- Aug 15, 2026
- Updated
- Aug 15, 2026