Hey HN, I'm Shreyash, founder of Feyn. We built Pulpie, a family of Pareto optimal models for cleaning the web. Pulpie strips boilerplate (ads, footers, sidebars) from raw HTML and returns just the main content as HTML or Markdown.

We match SOTA extraction quality while being 20x cheaper. Cleaning 1 billion webpages costs $7,900 with Pulpie versus $159,000 with Dripper, the current leading extractor.

The gains come from architecture. Today's leading extractors are decoders that generate output one token at a time. Each step reads the full model from memory to produce a single token. Conversely, Pulpie models are encoders. They run one forward pass over the full input HTML and label each block as boilerplate or content. As a result, Pulpie is compute-bound while decoders are memory-bound. Cheaper GPUs have relatively more compute than memory bandwidth. This makes Pulpie easy to run optimally.

Here's Pulpie and Dripper cleaning the same pages side by side: https://www.youtube.com/watch?v=ibd-tIiQECo. You can try a side-by-side comparison yourself: https://huggingface.co/spaces/feyninc/pulpie

Our motivation for Pulpie came from building a deep research harness. Every search API returns noisy content that contains ads, nav elements, and sidebars. In one instance, an ad for "Gemini on Pixel" slipped into our search results, got passed into LLM context, and ended up in the final answer served to the user. Pretty embarrassing moment for us but it helped us realize how bad data kills model intelligence. We built Pulpie to get clean data for cheap.

All models are open source on Hugging Face. You can read about our training process and how to use Pulpie here: https://usefeyn.com/blog/pulpie-pareto-optimal-models-for-cl...

Happy to answer any questions!

Why not use a plain old html → markdown converter? You can easily strip out ads using CSS /jQuery-like selectors. That would cost zero dollars.

It's good looking, and I liked it. The trial page accessed from the hugging face website is a very inefficient experience when I use Mozilla and the dark theme, FYI.

Leeches.

Very nice! Thank you for building this.

Why does the 'Quality vs Cost of Web Content Extraction' chart not have zero cost at the origin? Up to the right does not have to mean better; we can read.

ongrats on the release! The architectural insight to switch from a bandwidth-bound decoder (token-by-token generation) to a compute-bound encoder (single forward pass over 8k chunks) is brilliant—the 20x speedup and cost drop from $159k to $7.9k per billion pages is massive for web-scale pipelines.

As someone building AI developer utilities and document tools, I have two quick technical questions:

How well does the <|sep|> block-marker architecture handle heavily obfuscated HTML or adversarial SEO spam where boilerplate is styled to look like semantic body text? Have you tested running pulpie-orange-small (210M) quantized (e.g., INT8/FP8) on consumer edge GPUs or CPU-only setups for local RAG pipelines? Amazing work on open-sourcing the teacher and distilled weights on Hugging Face!