Invoice and form data extraction: what it costs and which AI to use
Pull fields such as supplier, date, totals and line items out of invoices, receipts and forms into JSON. Finance teams, accountants and software vendors processing paperwork at volume.
The short answer
At 20,000 documents a month, we estimate $4.65 on DeepSeek V4 Flash 0423 (the cheapest model that cleared our quality bar), $49.85 on Kimi K2.7 Code, and $416 on GPT-5.5. The flagship costs 8.3x the middle pick. Judged on JSON-extraction tasks: valid structure and correct values, checked automatically.
Simulate your invoice and form data extraction
Pick a job, set how much of it you do, and see what every model would cost. Defaults are typical, not yours. Change them.
Prompt, history, documents
What the model writes
Advanced: caching, reasoning, retries, discounts
Share that can use cached-input prices where a model offers them
1 = none. 2 = hidden thinking doubles billed output
Extra calls on top of the base count
Batch jobs or committed spend. Check your provider; we assume none by default
Lowest benchmark score a model can have and still be picked. Higher is stricter.
Limit to certain providers
Untick any provider you cannot use, for data-residency, contract or policy reasons. Picks are recalculated.
20K documents a month, three ways to pay
Every model here cleared a score of 9.0 out of 10 on the extract json benchmark tasks, ran without errors and fits a 16K-token window. 29 of 386 priced models qualify.
Cheapest model that clears the bar.
DeepSeek V4 Flash 0423
- Per document
- $0.0002
- Benchmark
- 9.67 / 10 · 3 prompts
- Context
- 1.05M tokens
Priced near the middle of the models that clear the bar; best score there.
Kimi K2.7 Code
- Per document
- $0.0025
- Benchmark
- 9.67 / 10 · 3 prompts
- Context
- 262K tokens
What many people reach for first.
GPT-5.5
- Per document
- $0.021
- Benchmark
- 9.67 / 10 · 3 prompts
- Context
- 1.05M tokens
Choosing GPT-5.5 over Kimi K2.7 Code costs $366 more a month, $4,395 a year (8.3x the price).
Both scored 9.67 out of 10 on this kind of work. The extra money buys safety margin our benchmark cannot measure: it is 24 short prompts, and most passing models score near the top. If a wrong answer is costly for you, that margin can be worth it. Test on your own data before you decide.
If it grows
Pay-per-use cost scales in a straight line. Monthly cost at 1x, 3x, 10x and 100x today's volume.
| Model | 1x · 20K | 3x · 60K | 10x · 200K | 100x · 2M |
|---|---|---|---|---|
| DeepSeek V4 Flash 0423 | $4.65 | $13.95 | $46.49 | $465 |
| Kimi K2.7 Code | $49.85 | $150 | $499 | $4,985 |
| GPT-5.5 | $416 | $1,248 | $4,161 | $41,612 |
What a budget buys
- DeepSeek V4 Flash 0423430Kdocuments a month, about 14K a day
- Kimi K2.7 Code40Kdocuments a month, about 1,337 a day
- GPT-5.54,806documents a month, about 160 a day
Every model, priced for your workload
41 models, 11 hidden because their window is smaller than 16K tokens.
| Model | A month | Per document | Benchmark | Context |
|---|---|---|---|---|
| DeepSeek V4 Flash 0423 DeepSeekBudget pick | $4.65 | $0.0002 | 9.67 (3) | 1.05M |
| DeepSeek V4.1 Flash DeepSeek | $5.75 | $0.0003 | 9.67 (3) | 1.05M |
| Llama 4 Scout Meta | $6.22 | $0.0003 | 9.67 (3) | 1.31M |
| Ministral 3 8B 2512 Mistral AI | $6.30 | $0.0003 | 9.67 (3) | 262K |
| Qwen3.8 Flash Alibababelow the bar | $8.95 | $0.0004 | 7.67 (3) | 1M |
| GLM 5.3 Flash Z Ai | $9.31 | $0.0005 | 9.67 (3) | 1.05M |
| Mistral Small 4 Mistral AI | $10.01 | $0.0005 | 9.67 (3) | 262K |
| GLM 5.3 Z Aibelow the bar | $14.41 | $0.0007 | 7.67 (3) | 1.05M |
| MiniMax M2.5 Minimax | $16.97 | $0.0008 | 9.67 (3) | 205K |
| GPT-5.4 Nano OpenAI | $17.06 | $0.0009 | 9.67 (3) | 400K |
| Muse Glimmer 30B Meta | $20.11 | $0.0010 | 9.67 (3) | 131K |
| MiniMax M2.7 Minimaxbelow the bar | $20.27 | $0.0010 | 6.67 (3) | 205K |
Benchmark = mean judge score out of 10 on the extract json prompts of the Model Playground, with the number of prompts in brackets. Models marked n/a have no benchmark run, so we do not guess their quality. How the benchmark works.
What we assumed
- Twenty thousand documents a month.
- Each document is about 1,300 words of extracted text plus a schema, around 2,000 tokens.
- The JSON that comes back is about 400 tokens.
- 3% of calls are retried after a failed parse.
These are typical numbers, not measurements of your business. Edit them in the simulator above.
What drives the cost
Volume. The documents are short, but there are thousands, and a few percent failed parses force expensive retries.
Watch out for
- This estimate assumes text is already extracted. Scanned pages that need a vision model cost more per page.
- Many providers discount batch jobs. If results can wait a few hours, use the discount slider.
Questions people ask
How much will invoice and form data extraction cost per month?
At 20,000 documents a month, we estimate $4.65 a month on DeepSeek V4 Flash 0423, the cheapest model that clears our quality bar for this job. Kimi K2.7 Code costs $49.85 and GPT-5.5 costs $416. These are estimates from the assumptions listed on this page, not a quote. Change the numbers to match your own volume.
What is the cheapest AI model for invoice and form data extraction?
On current OpenRouter prices, DeepSeek V4 Flash 0423 from DeepSeek is the cheapest model that cleared our bar of 9.0 out of 10, scoring 9.7/10 on 3 benchmark prompts for this kind of work. The benchmark is small, so treat it as a shortlist and test on your own data before switching.
Is it worth paying for a flagship model for invoice and form data extraction?
On our benchmark, GPT-5.5 scored 9.7/10 on this kind of work, and a model priced near the middle of the models that cleared the bar, Kimi K2.7 Code, scored 9.7/10, while the flagship costs 8.3x as much. A flagship buys safety margin the benchmark cannot measure. If a wrong answer is expensive for you, that can be worth paying for. If not, you are probably overpaying.
What happens to the cost if invoice and form data extraction grows ten times?
Pay-per-use pricing scales in a straight line, so ten times the documents is ten times the bill. Use the growth table on this page to see 3x, 10x and 100x for the same three models.
What makes invoice and form data extraction expensive?
Volume. The documents are short, but there are thousands, and a few percent failed parses force expensive retries.
Price a different job
- Customer support chatbotFrom $1.74 a month at 3,000 conversations
- Coding assistant for a teamFrom $4.12 a month at 4,400 requests
- Document summarisationFrom $1.86 a month at 2,000 documents
- Blog and content writingFrom $0.083 a month at 120 articles
- Knowledge base Q&A (RAG)From $5.62 a month at 15K questions
- Ticket and content classificationFrom $11.30 a month at 500K items
- AI agent workflowsFrom $7.44 a month at 1,500 runs
- Sales email personalisationFrom $0.595 a month at 5,000 emails
- Meeting notes and summariesFrom $0.575 a month at 400 meetings
- Code review for pull requestsFrom $0.840 a month at 600 pull requests
- Personal AI assistant for a founderFrom $0.478 a month at 1,500 messages
- Text-to-SQL data analystFrom $2.54 a month at 6,000 queries