Pulpie Orange Small extracts the main content from raw HTML, stripping navigation, ads, sidebars, and footers. It is an encoder that labels every HTML block as content or boilerplate in a single forward pass, so it approaches state-of-the-art extraction quality while running far faster and cheaper than autoregressive extractors.
At 210M parameters it matches Dripper (0.6B) in quality on WebMainBench (0.862 vs 0.864) while running 20x faster on an L4 GPU. It has the best size-to-quality ratio in the Pulpie family and is the recommended model for production.
Usage
The easiest way to use this model is through the pulpie package, which handles HTML simplification, chunking, classification, and reconstruction:
pip install pulpie
from pulpie import Extractor
extractor = Extractor() # defaults to pulpie-orange-small
result = extractor.extract(html)
print(result.markdown) # clean Markdown
print(result.html) # clean HTML
print(result.n_main, result.n_other) # blocks kept vs dropped
Extractor auto-detects CUDA, Apple MPS, then CPU. See the GitHub README for batch and multi-GPU usage.
How it works
Pulpie runs a four-stage pipeline:
- Simplify — remove scripts, styles, and formatting noise; tag each block with a unique ID.
- Chunk — pack blocks into sequences of up to 8,192 tokens separated by
<|sep|>markers (~80% of pages fit in one chunk). - Classify — a single encoder forward pass labels every block (at its
<|sep|>position) as content or boilerplate. - Reconstruct — return the kept blocks as HTML, or convert to Markdown.
This model is a token-classification head over EuroBERT-210m, distilled from the 2.1B Pulpie Orange Large teacher (KL-divergence 0.7 + hard-label cross-entropy 0.3, temperature 2.0).
Benchmarks
WebMainBench, English subset (6,647 pages), ROUGE-5 F1:
| Model | Params | ROUGE-5 F1 | Throughput (L4) |
|---|---|---|---|
| Pulpie Orange Large | 2.1B | 0.873 | 1.3 pages/sec |
| Dripper | 0.6B | 0.864 | 0.68 pages/sec |
| Pulpie Orange Base | 610M | 0.863 | 3.9 pages/sec |
| Pulpie Orange Small (this model) | 210M | 0.862 | 13.7 pages/sec |
| magic-html | - | 0.700 | - |
| Trafilatura | - | 0.619 | - |
Cleaning 1 billion pages on an L4 costs ~$7,900 with Pulpie Orange Small versus ~$159,000 with Dripper. Full analysis in the blog post.
Model family
| Model | Params | ROUGE-5 F1 | Use case |
|---|---|---|---|
| pulpie-orange-small | 210M | 0.862 | Recommended — best value, fastest |
| pulpie-orange-base | 610M | 0.863 | Balanced |
| pulpie-orange-large | 2.1B | 0.873 | Highest quality (teacher) |
Acknowledgements
Pulpie builds directly on the work of the MinerU-HTML and Dripper team (Ma et al., 2025). Their simplify_html preprocessing, block-level annotation scheme, and the WebMainBench benchmark are foundational to this work. Built on EuroBERT (Boizard et al., 2025).
Citation
@note{pulpie2026,
title = {Pulpie: Pareto-Optimal Models for Cleaning the Web},
author = {Minhas, Bhavnick and Nigam, Shreyash and Feyn Research},
year = {2026},
venue = {Feyn Field Notes}
}
Built by Feyn. Model weights licensed under CC BY-NC 4.0 (non-commercial); contact team@usefeyn.com for commercial licensing. The pulpie library is Apache 2.0.