Surya OCR 2 — MLX (mixed)

MLX conversion of datalab-to/surya-ocr-2 (a Qwen3.5/Qwen3-Next OCR VLM), runnable on Apple Silicon via mlx-vlm.

Variant mixed. See the project's bench/RESULTS.md for the precision benchmark. The mixed variant keeps the vision tower, embeddings, lm_head and the 18 Gated-DeltaNet layers at bf16 and quantizes only the bulk linears — strictly more precise than the existing 8-bit port.

from mlx_vlm import load, generate
model, processor = load("aglaia-models/surya-ocr-2-mlx")

Changes from the original

Weights converted to MLX format and quantized (mixed). No fine-tuning.

License & attribution

Derivative of datalab-to/surya-ocr-2, under the AI Pubs OpenRAIL-M License (Modified). This derivative is released under the same license (share-alike); the use-based restrictions carry over. Credit: Datalab (https://github.com/datalab-to/surya).

Downloads last month: 60

Safetensors

Model size

0.4B params

Tensor type

BF16

U32

MLX

Hardware compatibility

8-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for aglaia-models/surya-ocr-2-mlx

Base model

datalab-to/surya-ocr-2

Quantized

(6)

this model

Evaluation results

allenai/olmOCR-bench leaderboard
Overall View evaluation results

source

83.3
Arxiv Math View evaluation results

source

88.3
Old Scans Math View evaluation results

source

81.4