Roboflow

Best Document Question Answering Models

Compare the best 92 document question answering models and try 86 of them on your own images, free in the Roboflow Playground. 30 are open-weight, so you can self-host them for free under their licenses.

92 models · 30 open-weight · 86 free to try · prices synced Nov 18, 2026

We haven't benchmarked document question answering yet; scores and rankings appear on this site only where we've measured them. Until then, test the models below side by side, and for fixed categories in production expect a model fine-tuned on your own data to win.

Open-Source Document Question Answering Models

30 models with downloadable weights you can self-host under their licenses (Custom, MIT, and Apache 2.0). 27 run live in the Playground through hosted APIs, so self-hosting is optional.

Actions
816B total / 42B activeCustomOct 2026
Try
Mistral
1.05T total, 49B activeCustomOct 2026
Try
309B total, 15B activeMITSep 2026
Try
1.02T total, 42B activeMITSep 2026
Try
320B total, 18B activeMITAug 2026
Try
Qwen
27.78BApache 2.0Aug 2026
Try
29.6BApache 2.0Aug 2026
Try
MoonshotAI
Kimi K3Moonshot AI
2.8TModified MITJul 2026
Try

Document Question Answering Models via API

62 proprietary models where the weights aren't downloadable: access is through each provider's API and billed by them. Try 59 of them free in the Playground.

Actions
GoogleNEW
———Nov 2026
Try
AnthropicNEW
$8.00$40.001MNov 2026
Try
MetaNEW
$0.10$0.501.0MNov 2026
OpenAI
$1.00$5.001.1MOct 2026
Try
Google
$2.00$10.001MOct 2026
Try
Meta
$4.00$20.001.0MOct 2026
Try
Anthropic
$0.10$0.501MOct 2026
Try
OpenAI
$2.00$10.001.1MSep 2026
Try

How to Choose a Document Question Answering Model

Document QA layers reading, layout understanding, and reasoning. The practical choice is between end-to-end VLMs and structured extraction pipelines, and it turns on whether your questions are fixed or free-form.

If you always want the same fields (total, date, account number), frame it as structured extraction: prompt a document-capable model for a JSON schema, validate the output, and route failures to review. Fixed schemas are easier to evaluate and monitor than free-form answers, and cheaper models handle them reliably.

When users ask arbitrary questions about arbitrary documents, a strong VLM reading the page image end to end is the practical choice: it handles tables, key-value layouts, and reading order without a custom pipeline. Accuracy drops on dense tables and long multi-page documents, so keep humans in the loop where errors are costly.

The bottom line: Fixed fields deserve a schema-based extraction pipeline with validation; free-form questions justify an end-to-end VLM, with review on high-stakes answers.

Frequently Asked Questions About Document Question Answering Models

Yes. 30 of the 92 document question answering models here are open-weight (for example Muse Spark Open, Mistral Large 4, and MiMo V2.6 Flash), free to self-host under their licenses (Custom, MIT, and Apache 2.0).

Yes. You can run 86 of them in the Roboflow Playground for free. Upload an image and compare the models' output side by side, no setup required.

Explore other vision task pages, compare models head to head, or browse the full model catalog.