yoeven

Yup does really depend on the use case.

We see two types: workflows & agents.

Workflows are the most common, there's a pipeline like processing loan documents before data gets loaded to the next step or translating user comments before being stored in the database.

Agents are where you have a chat based system or a brain of sorts that calls many tools to achieve a user goal. The model doing this is a lot better at non deterministic task which then delegates to Interfaze for specific deterministic actions like OCR, Web extract then consumes that data. That's the article you referenced :)

Try it using run task mode when comparing to pure STT models: https://interfaze.ai/docs/audio/speech-to-text#long-audio-tr...

Excited to see the results

Thanks for the feedback!

We're working a lot more on speed in the coming few weeks :) More GPUs and more optimizations.

Our has been focus on quality of output first and we'll make optimizations as we grow :)

The lite models are great for simple use cases but won't don well in more complex OCR use cases.

yeah it would treat it like an OCR task and extract it, you could prompt it to format it better with the code alignment.

We serve it though an API. Check out the docs: https://interfaze.ai/docs

It's free to gets started.

It can, you could try prompting the model to use object detection vision and text extraction, we realized when we purely extract text it does amazing at word/sentence level bounds since the text acts as the anchor. However, when you treat it as a object detection problem, it sees that chunk of text as a segment allowing you the extract it as one column bound. Give that a try.

Not directly, LAMs tend to be focused a lot on tool calling or trained for a set of specific action for example in the robotics field. Good tool calling might be a good by product of Interfaze but wasn't specifically trained for that use case.

The focus has been for deterministic outputs that require high accuracy. In situations where there is "one right answer"

It wasn't designed to do well on MMMLU, it's a general model designed for deterministic task like OCR, object detection, STT and more and a by product of that is great language abilities. It still has a transformer backbone giving great language skills while being good at other stuff.

See the full benchmark: https://interfaze.ai/leaderboards

For sure there a tons of OCR bounding models and tons of other models like SAM 3 for segmentation.

Interfaze is a more powerful version of them combined into a single model, you can run multi turn tasks like extract all the text and object from this document then translate or generate a report.

It's like getting the best of both worlds from pure DNN/CNN models like Paddle and the flexibility and nuace of an LLM while outperforming both in accuracy.

Use it run task mode if you're doing a one to one comparison to whisper, it's going to be a lot faster too.

Here's a good example: https://interfaze.ai/docs/audio/speech-to-text#long-audio-tr...

The other way round, task specific DNNs adapted to share the same vector space as omni-transformers with generalized vision, audio encoders.

E.g. For an OCR task, the first pass will be handled by the CNN, converted to shared tokens which the transformer can consume, correct any issues if needed and a decoder that can handle both the DNN and transformer output.

It's a service API but we do have on prem deployment in certain regions for enterprises

Google Cloud Vision AI is a specialized model built on CNNs frameworks which is part of the Interfaze architecture which is an hybrid so you get best of both worlds. Google cloud vision was pretty far behind other specalized models like PaddleOCR etc anyways so if you're looking for a pure CNN, check them out.

You can find the explanation and the comparison in the article, which we benchmarked pure CNN models, pure LLM models and a hybrid architecture like ours.

Yup run task mode runs a much smaller part of the model when can drop quality of scans. The issue with run task we have to figure out is how much of the model is needed just for OCR and how to activate the right parts. A lot more improvements coming here with the same cost reduction.

I'd be happy to test it against your sample and see how we can get good results at a lower per page cost. Feel free to email me yoeven@interfaze.ai

It isn't on our roadmap right now since in most cases it should work out of the box and if it doesn't we'll work with you to train that into the model generally.

However, if we see enough people who has something super niche that our model can't handle, we might start considering a fine tuning service

Code extraction maybe, not something we have tested or built for but you could give it a try.

Code manipulation probably not since it's a lot smaller of a model compared to a Claude Opus which is SOTA for code generation/manipulation.

Generally code generation is a non-deterministic task by nature and general LLMs tend to be better at them.

We have a full benchmark breakdown specifically on structured output that you can take a look at https://interfaze.ai/leaderboards/structured-output-benchmar...

JigsawStack | Founding GTM (Go to market) / Growth | San Francisco, London, India, REMOTE (US) | Full Time

Company site: https://jigsawstack.com

At JigsawStack, we’re building specialized small models that automate complex backend tasks like OCR, web scraping, classification, data extraction, and more. We power thousands of developers and process billions of tokens every month.

We’re hiring our first Founding GTM, someone who wants to architect the entire GTM engine, experiment aggressively, and push the boundaries of AI growth.

You’ll own automation, outbound systems, GTM tooling, growth experiments, and the workflows that expand JigsawStack’s reach. This is a growth role where you’ll build pipelines, write scripts, automate everything possible, and systematically scale distribution.

As an early team member, you’ll work directly with the founders, run high-velocity experiments, and shape JigsawStack’s playbook for years to come.

More info and apply here: https://yoeven.notion.site/founding-gtm

Email here for more questions: careers@jigsawstack.com

Hey! Awesome product, not sure which model you use under the hood but you should check out https://jigsawstack.com/docs/api-reference/ai/vocr. Great for ID extraction which also includes bounding boxes and other attributes you'd need

Awesome stuff! This really helps

Can this be provided as an API?

It doesn't, it's a google translate widget alternative which is Open Source. Google deprecated their translate widget hence you can't really use google translate for this use case. So it's an open source framework that uses another translation service that can support this widget on the front end with the right security controls to prevent spam, etc.

The first comment explains the full story.

Hey HN! The easiest and fastest way to get a localized website was to add a single line for the Google Translate Widget script in your head tag.

Unfortunately, the Google widget was deprecated and has never been updated to match the modern standards of sites like React frameworks and many more. While the translation quality, especially, has lagged compared to where we are with AI today.

So I built the a open source translation widget that you can do a quick one script add and translate your website to 100+ languages. Comes with other fancy features like local storage caching, react page change watches, etc so you get the full package out of the box.

Yes next version will have over 98 languages :)

This is awesome, I've been following time series forecasting a ton so build a zero-shot time series model that allows you to pass in historical data and provide a 98% accurate forecast compared to custom training your own model with custom dataset. https://jigsawstack.com/ai-prediction

if you find it interesting, you could add it to your articles too as a comparison :)

Working link of image: https://www.reddit.com/media?url=https://preview.redd.it/vib...

We have a customer who has zero technical background and was vibe coding using Cursor. Basically hitting "Accept" all the way. At some point, Cursor suggested to install the JigsawStack SDK which then eventually prompted for an API key.

Naturally, he put his credit card down and got the API key on a $27/mth Pro Plan which includes 8m tokens of usage every month followed by $1.40 per million tokens.

A week later, he's on our support email surprised by a $200+ charge on his card. He was pissed, angry demanding a refund. Typically for situations like this, we do a partial refund to cover the cost for first-time customers if they honestly made a mistake and used more than intended.

That's when we realized, we didn't have a single $200+ charge tied to his account, only the $27/mth charge and we were so confused. We asked for the bank record/statement/invoice ID or anything that can help us find this charge.

That's when he sent an attached Cursor support email and invoices! And then all the pieces came together. He thought Cursor and JigsawStack were the same company because Cursor suggested to install the JigsawStack SDK. He got a $200+ charge from Vibe coding too hard! He was using max mode which is like 0.05 a prompt or something around those lines.

As you can see we have a language barrier as well. We tried our best to explain how these are two different companies and we weren't the ones who charged him. I think he got it since he's still a paying customer :)

Hey hey! Membership and Judging are the easiest items to get, and that's why i think your attorney really wants you hit that.

You don't have to join a accelerator where you have to give up equity. You could do something like ODF, or any program that has a selection process, is somewhat known and has a acceptance rate criteria. You could frame your VC as a membership too. For example YC is both a membership and a VC, Antler another. As long as your VC has a small membership page or program somewhere, you could frame it as such.

Judging is super simple, reach out to hackathons hosted by your friends, or in your area and ask if you can judge. Do a few, maybe 2-3. Try to get competitions where you co-judge are from companies with brand names like AWS, google, etc. That helps too.

How do you solve the access control part? Did you build the layer ontop or more like a auth redirect?

I think you can try the model

I ran Mistral AI OCR against JigsawStack OCR and beat their model in every category. Full breakdown here: https://jigsawstack.com/blog/mistral-ocr-vs-jigsawstack-vocr