fix: remove loguru

README.md

@@ -10,8 +10,6 @@ tags:
   - ocr
   - vision-language
   - document-intelligence
-base_model:
-  - deepseek-ai/DeepSeek-OCR
 inference: false
 model-index:
   - name: Jina-OCR-v1
@@ -54,6 +52,9 @@ This repository ships the model weights **and** the custom code needed to load t
 | ------------ | ----------------------------------------------------------- |
 | Backbone     | DeepSeek-OCR (DeepEncoder + 3B MoE, ~570M active)           |
 | Acceleration | FastMTP, one dense draft head, K=3 (vLLM only)              |
+| OmniDocBench v1.6 | 91.14 overall                                          |
+| olmOCR-Bench | 83.4 overall (+7.4 vs DeepSeek-OCR)                         |
+| Throughput   | 2.57 pages/s on olmOCR-Bench (A100, concurrency 32)         |
 | Local run    | [`example.py`](example.py) — Transformers or vLLM           |
 | Hosted       | OpenAI-compatible `https://api.jina.ai/v1/chat/completions` |
 
@@ -62,6 +63,24 @@ python example.py --backend transformers --image document.png
 python example.py --backend vllm --image document.png
 ```
 
+## Performance
+
+At the default dynamic-resolution setting, `jina-ocr-v1` scores **91.14** on OmniDocBench v1.6 and **83.4** on olmOCR-Bench, and has the highest page throughput of the fourteen systems we measured: **2.57 pages/s**. Post-training adds 7.4 points on olmOCR-Bench over the DeepSeek-OCR backbone. On an NVIDIA L4, FastMTP nearly doubles decoding speed over greedy autoregressive decoding. Scores below use the [eval prompt](#prompt).
+
+<p align="center">
+<img src="https://cms.jina.ai/content/images/size/w1600/2026/09/jina-ocr-v1-overview.png" alt="jina-ocr-v1 on three deployment-cost axes: pixels per visual token, page throughput, and accuracy versus active parameters">
+</p>
+
+Three axes that determine deployment cost. **(a)** Pixels per visual token, log scale. DeepEncoder maps a 1024×1024 view from 4,096 patches to 256 tokens (3,887 pixels per visual token vs 783–1,022 for 28–32 px patch encoders). **(b)** Page throughput on olmOCR-Bench, one A100, concurrency 32. **(c)** Benchmark overall vs active parameters (log scale); the solid line joins Pareto-optimal systems. `jina-ocr-v1` sits on both frontiers at 570M active parameters.
+
+<p align="center">
+<img src="https://cms.jina.ai/content/images/size/w1600/2026/09/jina-ocr-v1-efficiency.png" alt="olmOCR-Bench efficiency: output tokens per second, tokens per page, and pages per second">
+</p>
+
+The same fourteen systems on olmOCR-Bench (one A100, concurrency 32), ranked by **(a)** output tokens/s, **(b)** output tokens/page, and **(c)** pages/s. Surya OCR 2 leads on tokens/s (3,760) but emits 3,568 tokens/page and finishes 1.05 pages/s. `jina-ocr-v1` combines 2,792 tokens/s with 1,085 tokens/page and reaches 2.57 pages/s — the shortest outputs of any system scoring above 83.
+
+Per-category breakdowns on olmOCR-Bench and OmniDocBench v1.6, FastMTP measurements on L4, and the full comparison set are in the [technical report](https://arxiv.org/abs/2609.03181) and the [blog](https://jina.ai/news/jina-ocr-v1-faster-document-parsing-on-low-budget-gpus/).
+
 ## Prompt
 
 Recommended default:
@@ -193,13 +212,13 @@ Chat content types `image`, `image_pil`, and `image_url` all insert `<image>`. D
 
 ```bibtex
 @misc{garcía2026jinaocr,
-    title={Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards}, 
+    title={Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards},
     author={Alejandro Barón García and Feng Wang and Emilia Garcia Casademont and Han Xiao},
     year={2026},
     eprint={2609.03181},
     archivePrefix={arXiv},
     primaryClass={cs.CL},
-    url={https://arxiv.org/abs/2609.03181}, 
+    url={https://arxiv.org/abs/2609.03181},
 }
 ```
 

deepencoder.py

@@ -431,9 +431,6 @@ class VitModel(nn.Module):
         self.transformer = NoTPTransformer(cfg=cfg)
 
         if cfg.get("fp32norm", False):
-            from loguru import logger
-
-            logger.info("Load fp32 layernorm for ViT.")
             self.pre_layrnorm = LayerNormfp32(
                 cfg.hidden_size,
                 eps=cfg.get("pre_layernorm_epsilon", 1e-5),
@@ -1043,5 +1040,4 @@ def _build_sam(
         image_encoder.load_state_dict(
             {k[30:]: v for k, v in state_dict.items() if "vision_tower_high" in k}, strict=True
         )
-        print(checkpoint)
     return image_encoder

deepseek_ocr_mtp.py

@@ -491,8 +491,7 @@ def log_spec_stats(llm) -> None:
 
 
 DEFAULT_OCR_PROMPT = (
-    "Transcribe the provided document image into a clean Markdown format, "
-    "preserving the natural reading order."
+    "Transcribe the provided document image into a clean Markdown format, " "preserving the natural reading order."
 )
 
 

example.py

@@ -51,13 +51,7 @@ def run_vllm(
     max_new_tokens: int,
     num_speculative_tokens: int,
 ) -> str:
-    from deepseek_ocr_mtp import (
-        DEFAULT_OCR_PROMPT,
-        log_spec_stats,
-        register,
-        vllm_llm_kwargs,
-        vllm_sampling_params,
-    )
+    from deepseek_ocr_mtp import DEFAULT_OCR_PROMPT, log_spec_stats, register, vllm_llm_kwargs, vllm_sampling_params
     from vllm import LLM
 
     register()

modeling_deepseekocr.py

@@ -11,8 +11,8 @@ from transformers import LogitsProcessorList, TextStreamer
 from transformers.modeling_outputs import BaseModelOutputWithPast, MoeCausalLMOutputWithPast
 from transformers.utils import logging
 
-from .deepencoder import MlpProjector, _AttrDict, build_clip_l, build_sam_vit_b
 from .configuration_deepseek_v2 import DeepseekV2Config
+from .deepencoder import MlpProjector, _AttrDict, build_clip_l, build_sam_vit_b
 from .modeling_deepseekv2 import (
     DeepseekV2ForCausalLM,
     DeepseekV2Model,
@@ -444,9 +444,7 @@ class DeepseekOCRForCausalLM(DeepseekV2ForCausalLM):
             processors = LogitsProcessorList(processors)
         already = any(isinstance(p, SlidingWindowNoRepeatNgramProcessor) for p in processors)
         if ngram_size and window_size and not already:
-            processors.append(
-                SlidingWindowNoRepeatNgramProcessor(ngram_size, window_size, whitelist_token_ids)
-            )
+            processors.append(SlidingWindowNoRepeatNgramProcessor(ngram_size, window_size, whitelist_token_ids))
         if processors:
             kwargs["logits_processor"] = processors
         return super().generate(*args, **kwargs)

processing_deepseek_ocr.py

@@ -27,8 +27,7 @@ DS_OCR_TILE_SIZE = 640
 DS_OCR_BASE_SIZE = 1024
 
 DEFAULT_OCR_PROMPT = (
-    "Transcribe the provided document image into a clean Markdown format, "
-    "preserving the natural reading order."
+    "Transcribe the provided document image into a clean Markdown format, " "preserving the natural reading order."
 )
 
 DS_OCR_DEFAULT_CHAT_TEMPLATE = dedent(