[{"data":1,"prerenderedAt":415},["ShallowReactive",2],{"project-46006104":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":10,"maker":11,"posted_at":12,"year":13,"points":14,"comments":15,"rank_in_year":16,"shown_in_year":17,"top":18,"tier":19,"category":20,"subcategory":21,"format":22,"keywords":23,"reception":28,"sentiment":29,"stance_mix":30,"scored_comments":34,"aspects":35,"strip":44,"post_text":45,"what":46,"said":66,"maker_in_thread":191,"shape":201,"grid":205,"attempts":376,"same_week":377,"maker_shows":192,"maker_hidden":407,"reader_version":408,"card":409},"46006104","OCR Arena","A playground for OCR models","title","Show HN: OCR Arena – A playground for OCR models","https:\u002F\u002Fwww.ocrarena.ai\u002Fbattle","ocrarena.ai","kbyatnal",1763743485,2025,216,63,284,3837,8,"notable","ai","ocr model comparison","web_app",[24,25,26,27],"ocr","vision language model","ocr benchmark","document extraction","curious",-0.014,{"positive":31,"neutral":32,"mixed":33,"critical":31},5,36,3,49,{"pricing":36,"execution":38,"idea":39,"performance":42,"design":43},{"positive":37,"critical":37},0,{"positive":37,"critical":33},{"positive":40,"critical":41},2,1,{"positive":41,"critical":37},{"positive":37,"critical":41},[33,33,37,41,41,41,33,31,33,33,33,33,33,33,37,37,31,41,33,33,33,33,41,33,33,33,33,31,33,33,37,33,33,33,33,33,33,37,31,31],"I built OCR Arena as a free playground for the community to compare leading foundation VLMs and open-source OCR models side-by-side.\n\nUpload any doc, measure accuracy, and (optionally) vote for the models on a public leaderboard.\n\nIt currently has Gemini 3, dots.ocr, DeepSeek, GPT5, olmOCR 2, Qwen, and a few others. If there's any others you'd like included, let me know!",{"problem":47,"audience":53,"pricing":54,"motivation":58,"time_to_build":53,"stack":62,"openness":63,"stage":64,"team_size":65},{"value":48,"source":49,"evidence":50},"Hard to compare OCR models' accuracy on the same documents","maker",[51],{"id":52},"post",null,{"value":55,"source":49,"evidence":56},"Free to use",[57],{"id":52},{"value":59,"source":49,"evidence":60},"Community playground to compare leading foundation VLMs and open-source OCR models",[61],{"id":52},[],"free","launched","unknown",{"praise":67,"objections":84,"requests":118,"questions":169,"alternatives":177},[68],{"value":69,"source":70,"evidence":71},"Useful for comparing OCR models; a long-awaited benchmark","commenter",[72,76,80],{"id":73,"by":74,"text":75},"46006301","ylhert","We've got like 10 LLM arenas but nothing for OCR yet, really hope this takes off!",{"id":77,"by":78,"text":79},"46006984","profburial","This is a killer idea!",{"id":81,"by":82,"text":83},"46040721","zzleeper","Love this! Would have liked to see something like textract for a pre-LLM benchmark (but of course that's expensive), and also a distinction between handwritten text and printed…",[85,96,107],{"value":86,"source":70,"evidence":87},"Accuracy on complex tables is hard to judge; arenas are weak for correctness",[88,92],{"id":89,"by":90,"text":91},"46044660","molf","What is needed to evaluate OCR for most business applications (above everything else) is accuracy. Some results look plausible but are just plain wrong. That is worse than…",{"id":93,"by":94,"text":95},"46085599","fzysingularity","This is why arenas are generally a bad idea for assessing correctness in visual tasks.",{"value":97,"source":70,"evidence":98},"Bias toward English; no per-language split",[99,103],{"id":100,"by":101,"text":102},"46042312","deaux","I suggest you make explicit the assumption that this website is specifically about English text. Otherwise the leaderboard is pretty meaningless, with extreme differences in…",{"id":104,"by":105,"text":106},"46042770","skissane","LMArena splits their leaderboard by language: maybe you should consider doing the same thing I assume to do that you’d need another model to do language detection on the inputs…",{"value":108,"source":70,"evidence":109},"Some models hang or loop, and the UI gets stuck",[110,114],{"id":111,"by":112,"text":113},"46041757","daemonologist","A limitation of this leaderboard approach that I want to point out is that while the large general-purpose LLMs can make greater leaps of inference (on handwriting and poor…",{"id":115,"by":116,"text":117},"46042702","ajmurmann","Really like the idea. Unfortunately, my first upload is still spinning on one of the models about 5 minutes in. Clicking \"Stop Battle\" seems to do nothing either",[119,142,153,162],{"value":120,"source":70,"evidence":121},"Add more OCR models such as Mistral, Paddle, Chandra, Landing.ai DPT-2",[122,126,130,134,138],{"id":123,"by":124,"text":125},"46041083","codeddesign","Most of these are general LLM’s and not specifically OCR models. Where is Google Vision, Mistral, Paddle, Nanonets, or Chandra??",{"id":127,"by":128,"text":129},"46040631","ianhawes","Please add Chandra by Datalab",{"id":131,"by":132,"text":133},"46043724","densekernel","Any plans to add Document Pre-trained transformer-2 (DPT-2) from https:\u002F\u002Flanding.ai\u002F?",{"id":135,"by":136,"text":137},"46042659","coulix","We need to see Landing.ai DPT-2, from my tests its the best in term of ability to extract structure from complex tables so far.",{"id":139,"by":140,"text":141},"46046029","timbmg","Would be great to add MistralOCR!",{"value":143,"source":70,"evidence":144},"Mobile UX fixes and MathJax rendering",[145,149],{"id":146,"by":147,"text":148},"46042316","ComputerGuru","Two suggestions: UX on mobile isn’t great. It wasn’t obvious to me where the second model output was and I was thrown off even more so because the option to vote for model 1…",{"id":150,"by":151,"text":152},"46042721","rubikscubeguy","Hey! Dev who made this here. I hear you on the mobile UX, it's on my docket of things to fix. Same with math plugin! Thanks for the suggestions.",{"value":154,"source":70,"evidence":155},"Latency and cost metrics alongside quality",[156,159],{"id":157,"by":94,"text":158},"46040635","FYI one of the models on the battle was pretty slow to load. Are these also being rated on latency or just quality?",{"id":160,"by":11,"text":161},"46041260","Ultimately, there’s some intersection of accuracy x cost x speed that’s ideal, which can be different per use case. We’ll surface all of those metrics shortly so that you can…",{"value":163,"source":70,"evidence":164},"Programmatic API access",[165],{"id":166,"by":167,"text":168},"46045792","deeptishukla22","Is there a way I can invoke this programatically?",[170],{"value":171,"source":70,"evidence":172},"Will traditional OCR tools like Tesseract be added?",[173],{"id":174,"by":175,"text":176},"46041685","cdrini","There have been such a large number of OCR tools pop up over the past ~year; sorely in need for some benchmarks to compare them. Would love to see support for normal OCR tools…",[178],{"value":179,"source":70,"evidence":180},"Chandra, MinerU, Azure Document Intelligence, Landing.ai DPT-2",[181,182,186,190],{"id":127,"by":128,"text":129},{"id":183,"by":184,"text":185},"46049124","vdm","cool UI and lets anyone upload a doc. but lacks https:\u002F\u002Fgithub.com\u002Fopendatalab\u002Fmineru",{"id":187,"by":188,"text":189},"46010072","athoscouto","Nice! Would love to see Azure Document Intelligence on this",{"id":135,"by":136,"text":137},{"comments":192,"answered":40,"key":193},4,[194,197,198],{"id":195,"text":196},"46041184","We wanted to keep the focus on (1) foundation VLMs and (2) open source OCR models. We had Mistral previously but had to remove it because their hosted API for OCR was super…",{"id":160,"text":161},{"id":199,"text":200},"46041162","Sonnet\u002FOpus is being added shortly!",{"commenters":202,"max_depth":33,"maker_replies":192,"first_comment_mins":203,"lifespan_hours":204},46,20,182,[206,210,214,217,221,224,226,228,229,233,235,237,241,243,244,245,247,250,251,252,255,257,261,265,268,271,273,278,281,283,287,291,296,298,300,302,304,306,308,313,315,319,321,322,325,327,330,334,335,337,341,344,348,350,352,356,360,362,364,368,372,374,375],[207,37,33,208,209,37],"46040884","project\u002Fneutral\u002Fpricing","ArcaneMoose",[211,37,37,212,213,37],"46041899","commenter\u002Fneutral\u002Fnone","irjustin",[215,37,33,216,151,37],"46042712","maker\u002Fneutral\u002Fnone",[218,37,37,219,220,37],"46042858","other\u002Fneutral\u002Fnone","nicman23",[222,37,37,219,223,37],"46051655","intalentive",[89,41,41,225,90,37],"project\u002Fcritical\u002Fexecution",[93,41,41,227,94,37],"project\u002Fcritical\u002Fidea",[100,40,41,225,101,37],[230,40,33,231,232,37],"46042622","project\u002Fneutral\u002Fnone","hdjrudni",[234,40,33,231,101,37],"46042906",[236,40,37,219,220,37],"46042988",[238,40,31,239,240,37],"46045068","project\u002Fpositive\u002Fperformance","maverwa",[242,40,33,216,151,37],"46042720",[104,40,33,231,105,37],[174,33,33,231,175,37],[111,192,33,246,112,37],"project\u002Fmixed\u002Fexecution",[248,192,33,249,151,37],"46041793","maker\u002Fneutral\u002Fexecution",[123,31,33,231,124,37],[195,31,37,216,11,41],[253,31,33,254,140,37],"46046126","project\u002Fneutral\u002Fexecution",[256,31,33,231,151,37],"46041530",[258,259,33,246,260,37],"46044982",6,"poulpy123",[262,263,37,219,264,37],"46042900",7,"est",[266,263,37,219,267,37],"46043720","yorwba",[269,263,37,219,270,37],"46049039","tensor",[81,18,31,272,82,37],"project\u002Fpositive\u002Fnone",[274,275,33,276,277,37],"46044149",9,"project\u002Fneutral\u002Fperformance","tarruda",[146,279,41,280,147,37],10,"project\u002Fcritical\u002Fdesign",[150,279,33,282,151,37],"maker\u002Fneutral\u002Fdesign",[284,285,33,231,286,37],"46041979",11,"hakunin",[288,289,33,231,290,37],"46045879",12,"tethys",[292,293,33,294,295,37],"46041525",13,"project\u002Fneutral\u002Fidea","wener",[297,293,37,212,151,37],"46041538",[127,299,33,231,128,37],14,[115,301,41,225,116,37],15,[303,301,33,249,151,37],"46042711",[131,305,33,231,132,37],16,[135,307,33,231,136,37],17,[309,310,33,311,312,37],"46042231",18,"project\u002Fneutral\u002Fdesign","prodigycorp",[157,314,33,276,94,37],19,[316,314,33,317,318,37],"46041195","maker\u002Fneutral\u002Fperformance","andrewlu0",[320,314,33,276,232,37],"46042628",[160,314,37,317,11,41],[323,203,31,272,324,37],"46050068","mpercy123",[139,326,33,231,140,37],21,[328,326,33,208,329,37],"46048529","dahateb",[331,332,33,231,333,37],"46040758",22,"krashidov",[199,332,37,216,11,41],[336,332,33,231,151,37],"46041527",[338,339,33,231,340,37],"46047380",23,"gfody",[183,342,33,343,184,37],24,"project\u002Fmixed\u002Fnone",[345,346,33,231,347,37],"46042437",25,"mkolodny",[349,346,33,216,151,37],"46042722",[166,351,33,231,167,37],26,[353,354,33,231,355,37],"46047539",27,"lokl",[357,358,33,231,359,37],"46040370",28,"arathis",[361,358,37,216,11,41],"46041185",[363,358,33,231,151,37],"46041532",[365,366,37,219,367,37],"46043174",29,"aixpert",[369,370,37,219,371,37],"46040245",30,"dang",[73,370,31,373,74,37],"project\u002Fpositive\u002Fidea",[77,370,31,373,78,37],[187,370,33,231,188,37],[],[378,382,387,392,397,402],{"id":379,"name":380,"points":381,"category":20},"46036908","HN Simulator",538,{"id":383,"name":384,"points":385,"category":386},"46017175","Forty.News",443,"productivity",{"id":388,"name":389,"points":390,"category":391},"46048252","Flowglad",405,"web_saas",{"id":393,"name":394,"points":395,"category":396},"46051449","KiDoom",362,"games",{"id":398,"name":399,"points":400,"category":401},"46037031","exposedbydefault",275,"security_privacy",{"id":403,"name":404,"points":405,"category":406},"46053076","Yandori News Flow",256,"data",false,"v2",{"id":410,"name":5,"line":6,"ai":407,"cat":20,"year":13,"points":14,"comments":15,"rec":28,"s":411,"top":18,"kw":412,"maker":11,"domain":10,"tier":19,"attempts":37,"strip":413,"url":9,"mix":414,"praise":69,"objection":86,"sub":21},46006104,-0.01,[24,25,26,27],"3301113533333300513333133335330333333055",[31,32,33,31],1793991134794]