Qwen-2.5-32B is now the best open source OCR model
Last week was big for open source LLMs. We got: - Qwen 2.5 VL (72b and 32b) - Gemma-3 (27b) - DeepSeek-v3-0324 And a couple weeks ago we got the new mistral-ocr model. We updated our OCR benchmark to include the new models. We evaluated 1,000 documents for JSON extraction accuracy. Major takeaways: - Qwen 2.5 VL (72b and 32b) are by far the most impressive. Both landed right around 75% accuracy (equivalent to GPT-4o’s performance). Qwen 72b was only 0.4% above 32b. Within the margin of error. - Both Qwen models passed mistral-ocr (72.2%), which is specifically trained for OCR. - Gemma-3…
Qwen-2.5-32B is now the best open source OCR model launched on April 1, 2025 with 211 votes, #167 of 7,818 launches that month and more than 95% of that year's launches. The site was still online in September 2026, 1 year on.
In plain words
Qwen 2.5 VL is an open-source optical character recognition model available in 32 billion and 72 billion parameter versions. Designed for document analysis and JSON data extraction, it achieves approximately 75% accuracy on OCR tasks, matching GPT-4o performance. The model is built for developers and organizations needing open-source alternatives to commercial OCR solutions. The benchmark dataset and evaluation code are publicly available for reproducibility and testing.
written from the facts on this page · September 2026
From the sources
OCR Benchmark. Contribute to getomni-ai/benchmark development by creating an account on GitHub.
A benchmarking tool that compares OCR and data extraction capabilities of different large multimodal models such as gpt-4o, evaluating both text and json extraction accuracy. The goal of this benchmark is to publish a comprehensive benchmark of OCRaccuracy across traditional OCR providers and multimodal Language Models. The evaluation dataset and methodologies are all Open Source, and we encourage expanding this benchmark to encompass any additional providers. The primary goal is to evaluate JSON extraction from documents. To evaluate this, the Omni benchmark runs Document ⇒ OCR ⇒ Extraction . Measuring how well a model can OCR a page, and return that content in a format that an LLM can…from github.com
In the maker’s words, at launch
Last week was big for open source LLMs. We got: - Qwen 2.5 VL (72b and 32b) - Gemma-3 (27b) - DeepSeek-v3-0324 And a couple weeks ago we got the new mistral-ocr model. We updated our OCR benchmark to include the new models. We evaluated 1,000 documents for JSON extraction accuracy. Major takeaways: - Qwen 2.5 VL (72b…
Does a similar job
all alternatives →
Qwen3.5 SmallMar 2026 · huggingface.co · ▲3550.8B-9B native multimodal w/ more intelligence, less compute
Qwen-Image-2512Jan 2026 · qwen.ai · ▲316This is an independent project and is not affiliated with or endorsed by Qwen or Alibaba
LLM-aided OCR – Correcting Tesseract OCR errors with LLMs2024 · github.com · ▲479Almost exactly 1 year ago, I submitted something to HN about using Llama2 (which had just come out) to improve the output of Tesseract OCR by correcting obvious OCR errors [0]. That was exciting at the time because OpenAI's API calls were still quite expensive for GPT4, and the cost of running it on a book-length PDF would just be prohibitive. In contrast, you could run Llama2 locally on a machine with just a CPU, and it would be extremely slow, but "free" if you had a spare machine lying around. Well, it's amazing how things have changed since then. Not only have models gotten a lot better,…
More ai this month
the category →
Paintings by AI models: programs run through a simulation of oil paint.
AI · 22d ago · stillwet.art
FastRouter.ai▲364Route requests to the right LLM for cost, latency & quality
AI · 23d ago · fastrouter.ai
Source: Show HN post ↗ · launch text quoted in part
Launched alongside, April 2025
the whole month →
Hi everyone, I built PyXL — a hardware processor that executes a custom assembly generated from Python programs, without using a traditional interpreter or virtual machine. It compiles Python -> CPython Bytecode -> Instruction set designed for direct hardware execution. I’m sharing an early benchmark: a GPIO test where PyXL achieves a 480ns round-trip toggle — compared to 14-25 micro seconds on a MicroPython Pyboard - even though PyXL runs at a lower clock (100MHz vs. 168MHz). The design is stack-based, fully pipelined, and preserves Python's dynamic typing without static type restrictions.…
Dev tools · 2025 · runpyxl.com
The Unsure Calculator is an online tool that lets you calculate with numbers you’re not sure about.
Life & fun · 2025 · filiph.github.io
https://the-pocket.github.io/Tutorial-Codebase-Knowledge/
AI · 2025 · github.com
