Downloads · 30 days
17
18% of all-time downloads
chandra-ocr-2-oq4-mlx
1qh/chandra-ocr-2-oq4-mlx
chandra-ocr-2-oq4-mlx is a image-text-to-text model from 1qh. 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.
datalab-to/chandra-ocr-2 quantized with oQ (oMLX v0.5.3) at level 4.0, data-driven mixed precision: oQ measures each layer's quantization error through calibration and allocates bits where the measurement says they ma…
All-time downloads
93
Public
Parameters
4.5B
3.2 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors3.2 GB · 99%
Parameter types
How the weights are stored.
U324.2B · 93%
Try a prompt
Base models
Model card
From the Hugging Face model README
1qh/chandra-ocr-2-oq4-mlx
datalab-to/chandra-ocr-2 quantized with oQ (oMLX v0.5.3) at level 4.0,
data-driven mixed precision: oQ measures each layer's quantization error through calibration and
allocates bits where the measurement says they matter, rather than by a fixed per-tensor rule.
| Base | datalab-to/chandra-ocr-2 |
| oQ level | 4.0 (effective ~4.6–4.7 bits/weight) |
| Size | 3.2 GB (from 9.1 GB bf16) |
| Sensitivity entries | 248 |
| oMLX | v0.5.3 |
Table-column fidelity: FAILS. On the OCR fidelity check this build shifts config-table values by one column (every label carries the next column's value) while emitting well-formed output. Prefer level 5 or higher when table structure matters.
Vision weights are kept fp16; only the language tower is quantized.
Use
from mlx_vlm import load, generate
model, processor = load("1qh/chandra-ocr-2-oq4-mlx")
- Task
- Image-Text-to-Text
- Library
- mlx
- Type
- qwen3_5
- License
- apache-2.0
- Created
- Jul 24, 2026
- Updated
- Jul 24, 2026