What is the cheapest model for pulling data out of forms?

GLM-5.3-Flash at Merge Gateway, $0.028 per 1,000 forms on $0.01 / $0.05 per million tokens in and out, read on 16 Sept 2026.

How is the cost per unit worked out?

A scanned form is about 1,200 tokens in, images included, and 200 out. Multiply that by 1,000 and price it at each row's own input and output rate. Nothing else is counted: no cache discount, no batch rate, no free tier.

Why do only 130 models qualify?

The job needs 32k of context or more, structured output, and the ability to read images. Models that fall short are still in the index, they just cannot do this job as described.

How current are these prices?

Prices are read daily from 212 providers' public catalogs and this ranking recomputes with them. Last refresh: 16 Sept 2026.

Best AI model for pulling data out of forms.

prices as of 16 Sept 2026 · re-ranked daily · 130 qualifying models

Invoices, receipts and application forms arrive as scans, and what you want out of them is a handful of fields in a fixed shape. We count only models with 32k of context or more, structured output and the ability to read images, and 130 of the 337 we track qualify. The cheapest of them costs $0.028 per 1,000 forms at Merge Gateway. They are ranked by what the work costs, not by the price per million tokens, because a model that is cheap to prompt and dear to answer looks different once you know the shape of the job.

The cheapest model that can do this job is GLM-5.3-Flash at Merge Gateway: $0.028 per 1,000 forms, read on 16 Sept 2026.

Cheapest that qualifies

GLM-5.3-Flash

Merge Gateway

$0.028per 1,000 forms

$0.01 / $0.05 per M tokens · $0.14 at list

lowest cost per 1,000 forms

Cheapest from the maker

GPT-5.6 Luna Pro

OpenAI

$0.24per 1,000 forms

$0.10 / $0.60 per M tokens · $0.24 at list

bought from the lab that trained it

Most context

DeepSeek V4.1 Flash

TrustedRouter

$0.0675per 1,000 forms

$0.04 / $0.08 per M tokens · $0.30 at list

the longest single call on this list

What actually matters here

Reads images: Most forms arrive as a photograph or a PDF scan. A model that only takes text needs an OCR step in front of it, which is another bill and another thing to get wrong.
Structured output: You want fields, not a paragraph about the fields. These models can be held to a schema, so the answer drops straight into a database.
Input price decides it: An image is charged as input tokens, and six go in for every one that comes out.

Cheapest 10 for pulling data out of forms

1
Merge Gateway
GLM-5.3-Flash
$0.028 per 1,000 forms · $0.01 / $0.05 per M tokens · 1M contextHosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
✓ 1M context✓ tool calling✓ structured output
$0.10/M tokens
2
Eden AI
Qwen3.7 Flash
$0.0403 per 1,000 forms · $0.02 / $0.08 per M tokens · 1M contextAlibaba's China-region price list, which is lower than the international one. A region, not a deal.
✓ 1M context✓ tool calling✓ structured output
$0.14/M tokens
Eden AI ↗
3
Merge Gateway
Gemma 3 4B
$0.064 per 1,000 forms · $0.04 / $0.08 per M tokens · 131k contextHosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
✓ 131k context✓ structured output✓ reads images
$0.20/M tokens
4
TrustedRouter
DeepSeek V4.1 Flash
$0.0675 per 1,000 forms · $0.04 / $0.08 per M tokens · 1M contextRoutes to a host serving the open weights; precision not disclosed.
✓ 1M context✓ tool calling✓ structured output
$0.21/M tokens
TrustedRouter ↗
5
Crof
Qwen3.5-9B
$0.078 per 1,000 forms · $0.04 / $0.15 per M tokens · 262k contextHosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
✓ 262k context✓ tool calling✓ structured output
$0.27/M tokens
Crof ↗
6
Novita
Gemma 3 12B
$0.08 per 1,000 forms · $0.05 / $0.10 per M tokens · 131k contextThe same price OpenRouter charges.
✓ 131k context✓ tool calling✓ structured output
$0.25/M tokens
Novita ↗
7
Kilo Gateway
Gemma 4 26B A4B IT
$0.0944 per 1,000 forms · $0.04 / $0.22 per M tokens · 262k contextA promotion on Google's models: 50% off on Kilo's gateway, with no end date published.
✓ 262k context✓ tool calling✓ structured output
$0.35/M tokens
Kilo Gateway ↗
8
OpenRouter
Ling 3.0 Flash VL
$0.108 per 1,000 forms · $0.06 / $0.18 per M tokens · 131k contextThe same price OpenRouter charges.
✓ 131k context✓ tool calling✓ structured output
$0.36/M tokens
OpenRouter ↗
9
TrustedRouter
Qwen3.6 35B-A3B
$0.12 per 1,000 forms · $0.05 / $0.32 per M tokens · 262k contextRoutes to a host serving the open weights; precision not disclosed.
✓ 262k context✓ tool calling✓ structured output
$0.46/M tokens
TrustedRouter ↗
10
NanoGPT
Muse Spark 1.3 Contributor
$0.12 per 1,000 forms · $0.10 / $0.002 per M tokens · 1M contextThe same price Meta charges.
✓ 1M context✓ tool calling✓ structured output
$0.30/M tokens
NanoGPT ↗

A scanned form is about 1,200 tokens in, images included, and 200 out. A cost per 1,000 forms is that multiplied out at the row's own input and output rate. The price beside it is the standard price, three parts input to one part output per million tokens, so the two numbers answer different questions: $0.10 per million tokens is what the model costs, and $0.028 per 1,000 forms is what the work costs.

Frequently asked

What is the cheapest model for pulling data out of forms?: GLM-5.3-Flash at Merge Gateway, $0.028 per 1,000 forms on $0.01 / $0.05 per million tokens in and out, read on 16 Sept 2026.
How is the cost per unit worked out?: A scanned form is about 1,200 tokens in, images included, and 200 out. Multiply that by 1,000 and price it at each row's own input and output rate. Nothing else is counted: no cache discount, no batch rate, no free tier.
Why do only 130 models qualify?: The job needs 32k of context or more, structured output, and the ability to read images. Models that fall short are still in the index, they just cannot do this job as described.
How current are these prices?: Prices are read daily from 212 providers' public catalogs and this ranking recomputes with them. Last refresh: 16 Sept 2026.