Show HN: Free OCR Cost Calculator – Estimate multi-provider LLM token costs

A free LLM OCR pricing calculator published on a blog estimates that extracting a document with vision models costs between $0.0001 and $0.0160 per page, with GPT-6 Luna the cheapest and Claude Opus 5.5 the most expensive at roughly 43× the cost. At 10,000 receipts a month, the calculator puts monthly spend between $0.73 and $31.16, based on measured token counts from OpenAI, Google and Anthropic token counters and list prices checked 2026-10-27. The tool prices one call per document across three samples — an SROIE receipt, an AWS invoice and an IRS W-9 form — and lets users upload their own file, with input tokens ranging from 435 to 1,166 for a one-page receipt and 2,485 to 4,532 for a four-page invoice.

← Blog https://ocrbench.app/blog/ cost LLM OCR pricing calculator: your real cost per document Calculate the real cost per document and per month of LLM OCR. Compare the latest GPT, Gemini, Claude and Qwen models on measured token counts. Vision LLMs are priced in tokens, but you budget in documents and monthly bills. The gap between the two is where estimates go wrong, often by an order of magnitude: nobody knows offhand how many tokens a scanned invoice is, and the answer changes from one provider to the next. Short answer: LLM OCR costs $0.0001 to $0.0160 per document , depending on the model and the page. - Receipt: $0.0001 to $0.0031 GPT-6 Luna to Claude Opus 5.5 - W-9 form: $0.0001 to $0.0041 GPT-6 Luna to Claude Opus 5.5 - 4-page invoice: $0.0004 to $0.0160 GPT-6 Luna to Claude Opus 5.5 At 10,000 receipts a month that is $0.73 to $31.16 . The dearest model costs about 43× the cheapest, so the model you pick is the biggest cost lever. Prices checked 2026-10-27. This page closes that gap with measured numbers instead of guesses. Pick a sample document, or upload your own, and see what one extraction costs on each current model, then what it costs per month at your volume. The calculator Each row prices one call per document : the page image plus your prompt go in, and the JSON you expect comes out. Choose one of three measured samples a SROIE https://huggingface.co/datasets/jsdnrs/ICDAR2019-SROIE receipt, an AWS invoice, an IRS W‑9 https://www.irs.gov/pub/irs-pdf/fw9.pdf form , or upload your own file together with an example of the JSON you want back. Tick the models to price, then estimate. Set your documents per month, open any model row to see its arithmetic, and use "Copy link" to share the exact setup. At 10,000 receipts a month, LLM extraction costs $0.73 to $31.16 per month GPT-6 Luna is the cheapest and Claude Opus 5.5 the dearest, 43× apart. 1. input 544 tok × $0.10/M = $0.00005 + output 37 tok × $0.50/M = $0.00002 2. input 594 tok × $0.10/M = $0.00006 + output 37 tok × $0.50/M = $0.00002 3. input 435 tok × $0.15/M = $0.00007 + output 37 tok × $0.60/M = $0.00002 4. input 435 tok × $0.20/M = $0.00009 + output 37 tok × $0.88/M = $0.00003 5. input 544 tok × $2.00/M = $0.00109 + output 37 tok × $10.00/M = $0.00037 6. input 594 tok × $2.00/M = $0.00119 + output 37 tok × $10.00/M = $0.00037 7. input 1,166 tok × $2.00/M = $0.00233 + output 37 tok × $10.00/M = $0.00037 8. input 594 tok × $4.00/M = $0.00238 + output 37 tok × $20.00/M = $0.00074 Select a model to see the arithmetic. Costs are for one model call per document, at list prices. What is being priced Real token counts, no guessing. Your file goes to the provider's own token counter OpenAI, Google or Anthropic . No model reads it and nothing is stored. $0.0000 $0 input tokens × $ per 1M = cost + output tokens × $ per 1M = cost List prices checked 2026-10-27. Three things set your bill 1. The model. List prices differ by well over an order of magnitude between the cheapest and the most expensive current model, and that gap dominates everything below. 2. The document. Input is almost always the larger share of the cost: a page image becomes hundreds or thousands of tokens, while the extracted JSON is short. Page count scales it directly. A one-page receipt is 435–1,166 input tokens depending on the provider, a one-page W‑9 form is 631–1,166, and the four-page invoice is 2,485–4,532. 3. The provider's image handling. The same page costs a different number of tokens on each provider, because each one resizes and tiles images its own way. A cheaper per-token price does not guarantee a cheaper document, so compare cost per document, not price per million tokens. Cost per document at a glance The same numbers as the calculator, for every sample document and model, at 10,000 documents a month. | LLM OCR cost per document and per month at 10,000 documents a month. List prices checked 2026-10-27. | | | | | |---|---|---|---|---| | Model | Price per 1M tokens input / output | Receipt | W-9 form | 4-page invoice | |---|---|---|---|---| | GPT-6.1 Sol | $2.00 / $10.00 | $0.0015 $14.58 / month | $0.0019 $19.14 / month | $0.0075 $74.72 / month | | GPT-6 Luna | $0.10 / $0.50 | $0.0001 $0.73 / month | $0.0001 $0.96 / month | $0.0004 $3.74 / month | | Gemini 4 Argon | $2.00 / $10.00 | $0.0027 $27.02 / month | $0.0027 $26.72 / month | $0.0104 $104 / month | | Claude Haiku 5.5 | $0.10 / $0.50 | $0.0001 $0.78 / month | $0.0001 $1.03 / month | $0.0004 $4.00 / month | | Claude Sonnet 5.5 | $2.00 / $10.00 | $0.0016 $15.58 / month | $0.0021 $20.56 / month | $0.0080 $79.92 / month | | Claude Opus 5.5 | $4.00 / $20.00 | $0.0031 $31.16 / month | $0.0041 $41.12 / month | $0.0160 $160 / month | | Qwen3-VL 30B | $0.15 / $0.60 | $0.0001 $0.87 / month | $0.0001 $1.15 / month | $0.0005 $4.50 / month | | Qwen3-VL 235B | $0.20 / $0.88 | $0.0001 $1.20 / month | $0.0002 $1.56 / month | $0.0006 $6.11 / month | What that means in practice - One receipt costs $0.0001 with GPT-6 Luna and $0.0031 with Claude Opus 5.5, the two ends of the current range. - The four-page invoice costs $0.0004 to $0.0160. - At 100,000 documents , the receipt comes to $7.29 to $312, and the invoice to $37.36 to $1,598. How it is computed cost per document = input tokens × input price per 1M + output tokens × output price per 1M cost per month = cost per document × documents per month Each term is divided by 1,000,000 because prices are quoted per million tokens the receipt's GPT-6.1 Sol line is 544 × $2.00 ÷ 1,000,000 . That division is the step where most back-of-the-envelope estimates go wrong. Input tokens are the page s plus your prompt, counted by each provider's own free counter: OpenAI Responses input tokens , Gemini countTokens , Anthropic messages.count tokens . Qwen3‑VL has no free counter, so it is counted locally with its published image-packing rule. Models in the same generation share a sample count: both GPT-6 models, all three Claude 5.5 models, and both Qwen3-VL sizes. Those counts were measured on the sample page images with GPT-6.1 Sol, Gemini 4 Argon, and Claude Haiku 5.5. GPT-6 Luna, Claude Sonnet 5.5, and Claude Opus 5.5 return the same count on each of these pages, and both Qwen3-VL sizes use the same local rule. Output tokens are the JSON from your example, counted with the o200k base tokenizer for every model. The model's real answer will differ a little. Prices are catalogue list prices per million tokens. How to lower the bill - Pick the smallest model that passes your accuracy bar. This is the biggest lever by far, and it needs a benchmark on your own documents to answer. - Trim the prompt. It is sent with every document, so a long instruction block is multiplied by your volume. - Send only the pages you need. Drop cover pages, terms and blank sheets before the call. - Ask for only the fields you use. Output is the cheaper half, but a verbose schema still adds up at scale. - Check caching and batch options. Some providers discount cached input or asynchronous batch requests. The calculator does not apply these, so they come on top of the savings above. What the calculator leaves out - Retries and failures. Timeouts, rate limits and malformed JSON mean some documents are sent more than once. Measure your own retry rate and add it. - Caching and batch discounts. Not applied; see above. - Longer answers. A model that adds fields or explanations writes more output than your example. - More than one call. A classify step or an extract-then-verify pass means more model calls per document. Add the cost of each. - Per-page OCR services. Dedicated OCR APIs are typically priced per page rather than per token, so they are not in this calculator. Compare them on your own documents, on accuracy as well as price. Cost is one axis of three The cheapest model that gets the answer wrong is not cheap; it is a liability with a discount. Cost only means something next to accuracy and latency, so measure all three on the same documents before committing. See how to choose an OCR and extraction model https://ocrbench.app/blog/how-to-choose-an-ocr-extraction-model-2026/ for the full framework, and how OCR Bench scores extraction https://ocrbench.app/blog/how-ocr-bench-scores-extraction/ for how the accuracy half is graded. Use these numbers Everything above is free to reuse if you link back. Download the table as CSV https://ocrbench.app/blog/widgets/ocr-cost-calculator/ocr-cost-per-document.csv every document and model, with input and output tokens, prices and cost; the source is on every row . The Copy embed HTML button under the calculator puts it on your own page. The embedded version prices the three sample documents; the article is where you can upload your own. Frequently asked questions How much does LLM OCR cost per document? It depends on the model and the page. On the sample one-page receipt it costs $0.0001 with GPT-6 Luna and $0.0031 with Claude Opus 5.5, at the two ends of the current generation. The four-page invoice costs $0.0004 to $0.0160 depending on the model. How much does it cost to OCR 100,000 documents? At the receipt's per-document cost, 100,000 documents cost $7.29 with GPT-6 Luna and $312 with Claude Opus 5.5. For the four-page invoice the range is $37.36 to $1,598. Enter your own volume in the calculator for the exact figure. How many tokens is a scanned document? A one-page receipt is 435–1,166 input tokens and a one-page W-9 form is 631–1,166, depending on the provider. The four-page invoice is 2,485–4,532. The same image gets a different count from each provider because each resizes and tiles images by its own rules. Is LLM OCR cheaper than a traditional OCR API? Not necessarily. Dedicated OCR services are typically priced per page and LLMs per token, so the answer depends on page size, your prompt and the model. This calculator covers LLM token pricing only; to compare the options fairly, run the same documents through each and look at cost together with accuracy. Does the calculator include retries, caching or batch discounts? No. It prices one clean call per document at list prices. Retries add to the bill, while caching and batch APIs can lower it, so treat the result as a baseline. Is the document I upload stored? No. Your file is read in memory and sent to the provider's token counter to get the exact input count. No model runs on it and we do not save it. Measure cost and accuracy together Want the real per-document cost and accuracy for your documents, side by side? Start free https://ocrbench.app/register , run a benchmark, and read all three numbers off one scorecard.