{"slug": "glm-ocr", "name": "GLM-OCR", "category": "ai", "category_name": "AI tools and apps", "description": "", "recommenders": 6, "commenters": 7, "warners": 1, "score": 3.969, "rising": 0.69, "first": "2026-02-11", "latest": "2026-08-24", "why": "GLM-OCR is the best single model I have found for OCR over documents with good table extraction/formatting.", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 7, "neutral": 1, "compare": 0, "warn": 0}, "spark": [0, 0, 0, 2, 1, 1, 0, 0, 1, 2, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "GLM-OCR is the best single model I have found for OCR over documents with good table extraction/formatting.", "kind": "", "icon": true, "domain": null, "github_repo": "github.com/zai-org/glm-ocr", "official_url": "https://github.com/zai-org/GLM-OCR", "pricing": "", "aliases": ["GLM-OCR", "glm-ocr"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 790737409, "item_id": 49421088, "url": "https://news.ycombinator.com/item?id=49421088", "by": "phenomen", "date": "2026-08-24", "version": null, "stance": "recommend", "quote": "Another decent option is GLM OCR", "excerpt": "I tested many open-source and hosted OCR models and Datalab Chandra was the most accurate. It can parse complex layouts, tables, handwriting, and formulas at a fraction of the cost of Claude/Gemini. Local: https://github.com/datalab-to/chandra Hosted: https://www.datalab.to Another decent option is GLM OCR. It's slightly less accurate but faster and cheaper. Local: https://github.com/zai-org/GLM-OCR Hosted: https://d", "context": "", "use_case": "", "switched_from": "", "story": "OCR It \u2013 pull text out of un-copyable documents for your LLM"}, {"id": 789991540, "item_id": 49374471, "url": "https://news.ycombinator.com/item?id=49374471", "by": "cpnwaugha", "date": "2026-08-20", "version": null, "stance": "recommend", "quote": "you can switch between DeepSeek OCR 2, GLM-OCR, dots.mocr, Paddle OCR VL, PP-OCRv6, etc.", "excerpt": "Yeah, GPT 5.6 Sol is very good. Generally, Gemini models remain SOTA for VLM tasks with 3.7-flash at the top. --- That said, considering variables like cost (say, over 100k PDF pages) and accuracy requirements (e.g., construction documents with dense images & tables), Gemini and other SOTA VLMs are expensive and inaccurate, and therefore unsuitable. This is where niche, open-weight, and task-specific OCR/VL models co", "context": "", "use_case": "", "switched_from": "", "story": "GPT 5.6 Sol is the best \"vision\" model OpenAI ever released"}, {"id": 780009680, "item_id": 48750605, "url": "https://news.ycombinator.com/item?id=48750605", "by": "vinaigrette", "date": "2026-07-01", "version": null, "stance": "recommend", "quote": "I've had good results with glm-ocr on complex documents", "excerpt": "I'm surprised at the low rate every model manages considering the (apparent) ease of the benchmarked document. Can your pipeline produce ground truth as a byproduct ? How do you think open-weight ocr models compare to the one showcased ? I've had good results with glm-ocr on complex documents (complex by their handwriting, pretty easy layouts). What I like about your solution is the traceability of the information. A", "context": "", "use_case": "OCR of handwritten documents", "switched_from": "", "story": "Launch HN: Parsewise (YC P25) \u2013 Reason Across Documents with an API"}, {"id": 762615760, "item_id": 47663485, "url": "https://news.ycombinator.com/item?id=47663485", "by": "philipkglass", "date": "2026-04-06", "version": null, "stance": "recommend", "quote": "The one I have been recommending lately is GLM-OCR", "excerpt": "At least one of the test questions was just a screen shot from a tweet. It was hard to read. I'd suggest extracting text from screen shots with OCR. Apple has built-in functionality for this on their operating systems with Live Text. There are strong open source systems based on small vision language models for this, too. The one I have been recommending lately is GLM-OCR: https://github.com/zai-org/GLM-OCR It's fast", "context": "", "use_case": "OCR text extraction from images", "switched_from": "", "story": "Show HN: I Built Paul Graham's Intellectual Captcha Idea"}, {"id": 761919168, "item_id": 47619948, "url": "https://news.ycombinator.com/item?id=47619948", "by": "philipkglass", "date": "2026-04-02", "version": null, "stance": "recommend", "quote": "GLM-OCR is the best single model I have found for OCR over documents with good table extraction/formatting.", "excerpt": "If you have a basic ARM MacBook, GLM-OCR is the best single model I have found for OCR over documents with good table extraction/formatting. It's a compact 0.9b parameter model, so it'll run on systems with only 8 GB of RAM. https://github.com/zai-org/GLM-OCR Use the mlx-vlm back end: https://github.com/zai-org/GLM-OCR/blob/main/examples/mlx-deploy/README.md Then you can run a single command to process your PDF: glmo", "context": "", "use_case": "OCR of PDFs and tables", "switched_from": "", "story": "Google releases Gemma 4 open models"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 752879363, "item_id": 47054960, "url": "https://news.ycombinator.com/item?id=47054960", "by": "quinncom", "date": "2026-02-17", "version": null, "stance": "warn", "quote": "GLM-OCR didn't work because it returns all text on the screen", "excerpt": "I've got it installed with Qwen3-VL-4B running in LM Studio on my MBP M1 Pro. (Yes, the fans are running.) GLM-OCR didn't work because it returns all text on the screen, despite the instructions asking only for a summary. Screenshots are summarized in ~28 seconds. Here's the last one: > \"The user switched to the Hacker News tab, displaying item 47049307 with a \u201cGave Claude photographic memory for $0.0002/screenshot\u201d ", "context": "", "use_case": "screenshot summarization", "switched_from": "", "story": "Gave Claude photographic memory for $0.0002/screenshot"}]}], "who": [{"by": "philipkglass", "context": "former document-processing developer returning after a long break", "url": "https://news.ycombinator.com/item?id=47370066", "date": "2026-03-13"}], "tags": [], "use_cases": ["ocr of handwritten documents", "ocr text extraction from images", "ocr of pdfs and tables", "layout-aware document ocr", "screenshot summarization", "ocr model for screenshots"], "compared": [{"slug": "qwen", "name": "Qwen", "co": 3, "switched_from": 0, "switched_to": 0}, {"slug": "lm-studio", "name": "LM Studio", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=47049307", "title": "Gave Claude photographic memory for $0.0002/screenshot", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=49415852", "title": "OCR It \u2013 pull text out of un-copyable documents for your LLM", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49329575", "title": "GPT 5.6 Sol is the best \"vision\" model OpenAI ever released", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48746752", "title": "Launch HN: Parsewise (YC P25) \u2013 Reason Across Documents with an API", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=47663207", "title": "Show HN: I Built Paul Graham's Intellectual Captcha Idea", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=47616361", "title": "Google releases Gemma 4 open models", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {}, "2026-02": {"neutral": 1, "recommend": 2}, "2026-03": {"recommend": 1}, "2026-04": {"recommend": 1}, "2026-05": {}, "2026-06": {}, "2026-07": {"recommend": 1}, "2026-08": {"recommend": 2}, "2026-09": {}, "2026-10": {}}, "versions": [], "version_unnamed": 0, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 790737409, "item_id": 49421088, "url": "https://news.ycombinator.com/item?id=49421088", "by": "phenomen", "date": "2026-08-24", "version": null, "stance": "recommend", "quote": "Another decent option is GLM OCR", "excerpt": "I tested many open-source and hosted OCR models and Datalab Chandra was the most accurate. It can parse complex layouts, tables, handwriting, and formulas at a fraction of the cost of Claude/Gemini. Local: https://github.com/datalab-to/chandra Hosted: https://www.datalab.to Another decent option is GLM OCR. It's slightly less accurate but faster and cheaper. Local: https://github.com/zai-org/GLM-OCR Hosted: https://d", "context": "", "use_case": "", "switched_from": "", "story": "OCR It \u2013 pull text out of un-copyable documents for your LLM"}, {"id": 789991540, "item_id": 49374471, "url": "https://news.ycombinator.com/item?id=49374471", "by": "cpnwaugha", "date": "2026-08-20", "version": null, "stance": "recommend", "quote": "you can switch between DeepSeek OCR 2, GLM-OCR, dots.mocr, Paddle OCR VL, PP-OCRv6, etc.", "excerpt": "Yeah, GPT 5.6 Sol is very good. Generally, Gemini models remain SOTA for VLM tasks with 3.7-flash at the top. --- That said, considering variables like cost (say, over 100k PDF pages) and accuracy requirements (e.g., construction documents with dense images & tables), Gemini and other SOTA VLMs are expensive and inaccurate, and therefore unsuitable. This is where niche, open-weight, and task-specific OCR/VL models co", "context": "", "use_case": "", "switched_from": "", "story": "GPT 5.6 Sol is the best \"vision\" model OpenAI ever released"}, {"id": 780009680, "item_id": 48750605, "url": "https://news.ycombinator.com/item?id=48750605", "by": "vinaigrette", "date": "2026-07-01", "version": null, "stance": "recommend", "quote": "I've had good results with glm-ocr on complex documents", "excerpt": "I'm surprised at the low rate every model manages considering the (apparent) ease of the benchmarked document. Can your pipeline produce ground truth as a byproduct ? How do you think open-weight ocr models compare to the one showcased ? I've had good results with glm-ocr on complex documents (complex by their handwriting, pretty easy layouts). What I like about your solution is the traceability of the information. A", "context": "", "use_case": "OCR of handwritten documents", "switched_from": "", "story": "Launch HN: Parsewise (YC P25) \u2013 Reason Across Documents with an API"}, {"id": 762615760, "item_id": 47663485, "url": "https://news.ycombinator.com/item?id=47663485", "by": "philipkglass", "date": "2026-04-06", "version": null, "stance": "recommend", "quote": "The one I have been recommending lately is GLM-OCR", "excerpt": "At least one of the test questions was just a screen shot from a tweet. It was hard to read. I'd suggest extracting text from screen shots with OCR. Apple has built-in functionality for this on their operating systems with Live Text. There are strong open source systems based on small vision language models for this, too. The one I have been recommending lately is GLM-OCR: https://github.com/zai-org/GLM-OCR It's fast", "context": "", "use_case": "OCR text extraction from images", "switched_from": "", "story": "Show HN: I Built Paul Graham's Intellectual Captcha Idea"}, {"id": 761919168, "item_id": 47619948, "url": "https://news.ycombinator.com/item?id=47619948", "by": "philipkglass", "date": "2026-04-02", "version": null, "stance": "recommend", "quote": "GLM-OCR is the best single model I have found for OCR over documents with good table extraction/formatting.", "excerpt": "If you have a basic ARM MacBook, GLM-OCR is the best single model I have found for OCR over documents with good table extraction/formatting. It's a compact 0.9b parameter model, so it'll run on systems with only 8 GB of RAM. https://github.com/zai-org/GLM-OCR Use the mlx-vlm back end: https://github.com/zai-org/GLM-OCR/blob/main/examples/mlx-deploy/README.md Then you can run a single command to process your PDF: glmo", "context": "", "use_case": "OCR of PDFs and tables", "switched_from": "", "story": "Google releases Gemma 4 open models"}, {"id": 757921058, "item_id": 47370066, "url": "https://news.ycombinator.com/item?id=47370066", "by": "philipkglass", "date": "2026-03-13", "version": null, "stance": "recommend", "quote": "Try just GLM-OCR if you want to get started quickly", "excerpt": "I took a long break from document processing after working on it heavily 20 years ago. The tools I used before were ABBYY FineReader and PrimeOCR. I haven't tried any of the commercial cloud based solutions. I'm currently using GLM-OCR, Chandra OCR, and Apple's LiveText in conjunction with each other (plus custom code for glue functionality and downstream processing). Try just GLM-OCR if you want to get started quick", "context": "former document-processing developer returning after a long break", "use_case": "layout-aware document OCR", "switched_from": "", "story": "Can I Run AI locally?"}, {"id": 752895795, "item_id": 47055987, "url": "https://news.ycombinator.com/item?id=47055987", "by": "quinncom", "date": "2026-02-17", "version": null, "stance": "warn", "quote": "", "excerpt": "Update: I switched to Qwen3 VL 2B (`qwen3-vl-2b-instruct-mlx@bf16`) which is 2.5\u00d7 faster than 4B (11s vs 18s per screenshot) and my meager M1 Pro is able to keep up without the fans spinning 100% of the time.", "context": "", "use_case": "", "switched_from": "", "story": "Gave Claude photographic memory for $0.0002/screenshot"}, {"id": 752879363, "item_id": 47054960, "url": "https://news.ycombinator.com/item?id=47054960", "by": "quinncom", "date": "2026-02-17", "version": null, "stance": "warn", "quote": "GLM-OCR didn't work because it returns all text on the screen", "excerpt": "I've got it installed with Qwen3-VL-4B running in LM Studio on my MBP M1 Pro. (Yes, the fans are running.) GLM-OCR didn't work because it returns all text on the screen, despite the instructions asking only for a summary. Screenshots are summarized in ~28 seconds. Here's the last one: > \"The user switched to the Hacker News tab, displaying item 47049307 with a \u201cGave Claude photographic memory for $0.0002/screenshot\u201d ", "context": "", "use_case": "screenshot summarization", "switched_from": "", "story": "Gave Claude photographic memory for $0.0002/screenshot"}, {"id": 752848240, "item_id": 47053015, "url": "https://news.ycombinator.com/item?id=47053015", "by": "quinncom", "date": "2026-02-17", "version": null, "stance": "recommend", "quote": "GLM-OCR seems like it would excel at this", "excerpt": "Screen sharing to any remote API is a nonstarter for me. I don\u2019t care if the API claims ZDR; Snowden\u2019s revelations are still echoing. So, I appreciate that the app supports a custom endpoint for local models. Which local models did you try? GLM-OCR seems like it would excel at this: https://huggingface.co/zai-org/GLM-OCR", "context": "", "use_case": "OCR model for screenshots", "switched_from": "", "story": "Gave Claude photographic memory for $0.0002/screenshot"}, {"id": 752573312, "item_id": 47035832, "url": "https://news.ycombinator.com/item?id=47035832", "by": "xml", "date": "2026-02-16", "version": null, "stance": "recommend", "quote": "I've been using it for digitizing various hand-written notes", "excerpt": "You can still be excited! Recently, GLM-OCR was released, which is a relatively small OCR model (2.5GB unquantized) that can run on CPU with good quality. I've been using it for digitizing various hand-written notes all my shopping receipts this week. https://github.com/zai-org/GLM-OCR (Shameless plug: I also maintain a simplified version of GLM-OCR without dependency on the transformers library, which makes it much ", "context": "", "use_case": "OCR of handwritten notes and receipts", "switched_from": "", "story": "Rolling your own serverless OCR in 40 lines of code"}, {"id": 751631440, "item_id": 46976965, "url": "https://news.ycombinator.com/item?id=46976965", "by": "yieldcrv", "date": "2026-02-11", "version": null, "stance": "neutral", "quote": "", "excerpt": "> I burn up the entire context window for 30 pages of text We analyze 200 page contracts no problem I think you're doing it wrong or in an antiquated way (until context window sizes improve) Are you doing this programmatically at all or are you doing something closer to dropping a contract into a chat window? We use a main agent to classify the pages and we build subagents that are familiar with page classifications ", "context": "", "use_case": "OCR of scanned contracts", "switched_from": "", "story": "GLM-OCR: Accurate \u00d7 Fast \u00d7 Comprehensive"}]}