AIBuildAI ProWork-PDF-27B is Qwen3.8-27B post-trained by AIBuildAI on our own data for professional work on long documents: reading a working professional's request and the PDF packet it refers to (records, tables, charts, plans and regulations, often more than fifty pages), working out the figures from the packet, and answering in writing, as a memo, a note or a table.
We evaluate on GDP.pdf (Surge AI's 100 public tasks of document work across ten domains) under Artificial Analysis's protocol. The prompt carries the document's text and its page images. GPT-5.6 Luna (medium) judges every rubric criterion. All-pass is the share of attempts that meet every criterion.
Lighter bars are scores published by Artificial Analysis. All other bars are our own runs. We ran Qwen3.8-27B through its hosted API.
AIBuildAI ProWork-PDF-27B reaches 21.0% all-pass. That is 5.4 points above its base and 0.6 points below GPT-6 Luna (max).
Serve with vLLM:
vllm serve aibuildai-inc/ProWork-PDF-27B \
--max-model-len 262144 \
--reasoning-parser qwen3 \
--enable-auto-tool-choice --tool-call-parser qwen3_xml \
--speculative-config '{"method": "mtp", "num_speculative_tokens": 3}'
Then call it with any OpenAI-compatible client, giving the document's text and its page images:
from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")
response = client.chat.completions.create(
model="aibuildai-inc/ProWork-PDF-27B",
messages=[{"role": "user", "content": [
{"type": "text", "text": "Using the attached annual report, list each business segment with its revenue and operating margin, in a table. ..."},
{"type": "image_url", "image_url": {"url": "data:image/png;base64,..."}},
]}],
temperature=1.0,
top_p=0.95,
max_tokens=65536,
extra_body={"top_k": 20},
)
print(response.choices[0].message.content)
Recommended settings: thinking on, temperature 1.0, top-p 0.95, top-k 20, up to 65,536 output tokens per call. Give the model the document's text together with its page images, so it can read charts, plans and scanned tables.
As with any model, check figures, citations and legal or medical statements before relying on them.
Apache 2.0, the license of the base model Qwen3.8-27B.
@misc{aibuildai2026proworkpdf,
title = {AIBuildAI ProWork-PDF-27B},
author = {{AIBuildAI}},
year = {2026},
url = {https://huggingface.co/aibuildai-inc/ProWork-PDF-27B}
}
model-00004-of-00018.safetensors 4.0 GBmodel-00016-of-00018.safetensors 4.0 GBmodel-00006-of-00018.safetensors 4.0 GBmodel-00008-of-00018.safetensors 4.0 GBmodel-00010-of-00018.safetensors 4.0 GBmodel-00012-of-00018.safetensors 4.0 GBmodel-00014-of-00018.safetensors 4.0 GBmodel-00001-of-00018.safetensors 4.0 GBmodel-00018-of-00018.safetensors 3.4 GBmodel-00002-of-00018.safetensors 3.0 GBmodel-00003-of-00018.safetensors 2.5 GBmodel-00007-of-00018.safetensors 2.1 GBmodel-00009-of-00018.safetensors 2.1 GBmodel-00011-of-00018.safetensors 2.1 GBmodel-00013-of-00018.safetensors 2.1 GBmodel-00015-of-00018.safetensors 2.1 GBmodel-00017-of-00018.safetensors 2.1 GBmodel-00005-of-00018.safetensors 2.1 GBQwen/Qwen3.8-27B