# GUI Detection, #1 OCR Benchmark, Faster Reasoning & Smarter Scraping

URL: https://interfaze.ai/blog/20-04-26-updates

We launched on YC today: [YC Bookface](https://www.ycombinator.com/launches/Pyw-interfaze-deterministic-ai-for-developer-tasks-ocr-object-detection-web-scraping-stt-classification-and-more) | [X post](https://x.com/yoeven/status/2046274726121889879) | [Linkedin post](https://www.linkedin.com/posts/yoeven_launching-interfaze-with-yc-we-designed-ugcPost-7452041899865186304-elU0?utm_source=share&utm_medium=member_desktop&rcm=ACoAABYoZGIBzg2ab_Pks_mGonKQKmN6kId1mYo)

Here's what we shipped this week.

## GUI detection

Introducing GUI detection, natively detect buttons, icons, symbols, text, and more on any screen, like Mac, Linux, Windows and more. Works great for computer use applications that require high accuracy and speed when navigating.

GUI docs: https://interfaze.ai/docs/vision/gui-detection

![GUI detection](https://r2public.jigsawstack.com/interfaze/blogs/comouter_use_form.png)

_Filling out a form automatically_

## Reasoning quality improvement with fewer tokens

Reasoning is turned off by default but it can come in handy in some cases, like report generation, scientific calculations, or even a complex deep search.

Now, reasoning has improved performance, 2x faster and taking up to 50% fewer tokens to produce the same answer as before.

Reasoning docs: https://interfaze.ai/docs/reasoning

## OCR benchmark

We rank number one on OCR compared to all of the best in class SOTA models.

![OCRBench V2 benchmark](https://r2public.jigsawstack.com/interfaze/blogs/ocr_bench_v2_chart.png)

## Using interfaze as a tool inside your agent

Fun fact, many users use interfaze as deterministic tools within agentic workflows allowing their agent to delegate deterministic tasks like OCR or web scraping to interfaze.

Read full article: https://interfaze.ai/blog/using-interfaze-as-a-tool-inside-your-agent

![Using interfaze as a tool inside your agent](https://r2public.jigsawstack.com/interfaze/blogs/use_interfaze_as_a_tool_blog.png)

## Improved scraping for bot protected sites

Interfaze now has access to network logs from the scraping engine, allowing it to decrypt responses and find additional relevant data.

Scraping docs: https://interfaze.ai/docs/web/web-scraping

![Improved scraping for bot protected sites](https://r2public.jigsawstack.com/interfaze/blogs/scraping_linkedin_ex.png)

## Web search

Web search now includes 20 results on any run, which can be used for better context enrichment or post-processing.

Web search docs: https://interfaze.ai/docs/web/web-search

### Coming soon:

- New research paper on structured outputs
- Updated benchmark
- Performance improvements on caching

That's it for this week!

Best,
Yoeven
CEO, interfaze.ai
