This week, we launched Indra, a 1M parameter OCR model that's 15x faster than Tesseract on CPU and much more accurate.
I still can't believe how good this model is.
I still can't believe how good this model is.
Oct 16, 2026 · 5:42 PM UTC
109
242
2,942
104,887
@VikParuchuri Hi vik, it is indeed great! Faster and more accurate than Apple Vision on ANE on my machine. Any plans for additional finetuning on math/formulas/equations?
3
0
23
4,011
@VikParuchuri Hey Vik it works great. But on further tests it fails to recognize anything on slightly skewed pages like 7 desgress. On fully rotated pages however it works as its supposed to
1
0
8
2,244
@VikParuchuri If text is rotated, it fails to read. Imagine a text on Indian High School Certificate - Text in Semi Circle & Caligraphy, the output is not accurate on those text. How can we solve those cases?
1
0
2
1,802
@VikParuchuri What’s the Hebrew amount of data you have , and theoretically how much would you need of it more ?
1
0
2
1,776
@LyracoN_Peleg This model doesn't do hebrew, but future versions should. Feel free to DM if you have pointers to data!
2
0
1
1,499
@gauravprat This currently covers Latin, Cyrillic, Greek, Devanagari, Bengali, and Chinese.
0
0
12
1,630
@VikParuchuri Do you mean a 1M (post title) param model or 10M (chart number)?
1
0
0
1,377
@descentai The OCR model is 1.24M params, the tested config in the chart is with a special math model and a few others that are collectively 12M params
1
0
1
1,276
@VikParuchuri Amazing! Can't wait to put it through some tests. Does the devanagari recognize Vedic svaras?
0
0
0
225
@VikParuchuri It's very very good. So good I ported it to WASM and I'm running it in prod: github.com/RyanCasas/indr…
1
1
15
706
@VikParuchuri 100 万参数在 CPU 上跑赢 Tesseract 十几倍，这体积塞进手机端是不是毫无压力，我好奇准确率是在哪套语料上测的
0
0
1
130
@VikParuchuri Excited about this. Can’t wait to try it out.
0
0
1
75
@VikParuchuri There is no demo code on how to extract text from an independent image file using .py or .ipynb.
Image processing is only shown through CLI.
Image processing is only shown through CLI.
0
0
0
17
@VikParuchuri I implemented it in my ocr pipeline and iamreallysatisfied. I substituted ovis and get back nearly 10 gb of ram
0
0
0
189
@VikParuchuri does it work on scanned pdf, where there are actually images inside pdf files?
0
0
0
33
@VikParuchuri What would be the cost of parsing a medical report?
Also, are you guys compliant to process them?
Also, are you guys compliant to process them?
0
0
0
14
@VikParuchuri I will definitely check this out. I need something better and faster than glm-ocr.
0
0
0
75
@VikParuchuri Does this work when a photo is converted to PDF?
0
0
0
78
@VikParuchuri Can you try hard documents like those used in construction?
0
0
0
99
