---
title: "Most Token-Efficient Document Parsing API, Independently Benchmarked, 2026 | Openbenchmarks"
description: "Extend leads 7 specialised document parsers at 22.5M tokens per 1k correct, 93.1% accuracy on legal contracts. Independently benchmarked, 2026."
url: https://openbenchmarks.com/document-processing/most-token-efficient-document-parsing-api
---

[](https://openbenchmarks.com/) 

[Detailed benchmark →](https://openbenchmarks.com/document-processing) 

[benchmarks](https://openbenchmarks.com/) / [document processing](https://openbenchmarks.com/document-processing) / most token-efficient document parsing api 

independently benchmarked · 1,500 questions · 94 contracts · updated Oct 25, 2026

# Most Token-Efficient Document Parsing API, 2026

Extend leads 7 specialised document parsers at 22.5M tokens per 1k correct, 93.1% accuracy on legal contracts. Independently benchmarked, 2026.

**What this page compares.** Document parsers create the text used by downstream extraction, search, RAG, and agent workflows. This benchmark tests whether that text preserves the operative meaning of heavily redlined contracts: every system received the same complete PDF, and a fixed reader answered the same questions from each parser's Markdown.

**How to read the result.** The ranking uses every input and output token the same answering agent spent across all questions, divided by its correct answers, per 1,000\. Most of it is the parser's output, read again with every question, so a compact, faithful parse is the token-efficient one. A parser must beat the markup-blind baseline to be ranked, so it cannot win by returning less text.

**Workload and controls.** Each measured configuration processed the same 94 tagged contract PDFs. The same downstream reader then answered the same 1,500 questions. A best-case parse and a markup-blind baseline bound the score.

\[01\] measured ranking

## Ranked by the answering agent's total tokens per 1,000 correct answers

The ranking uses every input and output token the same answering agent spent across all questions, divided by its correct answers, per 1,000\. Most of it is the parser's output, read again with every question, so a compact, faithful parse is the token-efficient one. A parser must beat the markup-blind baseline to be ranked, so it cannot win by returning less text.

| Rank | Provider | Type | Accuracy | Gap closed | Median / P95 latency | Parser $ / 1k pages | Parser $ / 1k correct | Total tokens / 1k correct | List price |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | [Extend](https://docs.extend.ai/) engine=parse\_performance | Specialised document parsers | **93.1%** ± 0.3% | 97.9% | 17s / 26s | $20.00 | $32.73 | 22.5M | [$0.020 / page](https://www.extend.ai/pricing) |
| 2 | [LlamaParse agentic](https://docs.cloud.llamaindex.ai/llamaparse/overview) tier=agentic · 2026-09-29 | Specialised document parsers | **83.8%** ± 0.6% | 84.5% | 16s / 71s | $12.50 | $22.71 | 23.9M | [$0.0125 / page](https://developers.llamaindex.ai/llamaparse/general/pricing/) |
| 3 | [LlamaParse agentic plus](https://docs.cloud.llamaindex.ai/llamaparse/overview) tier=agentic\_plus · 2026-09-24 | Long running agentic parsers | **84.7%** ± 0.3% | 85.8% | 82s / 138s | $56.25 | $101.17 | 24.0M | [$0.05625 / page](https://developers.llamaindex.ai/llamaparse/general/pricing/) |
| 4 | [Datalab track changes](https://documentation.datalab.to/) track-changes endpoint | Specialised document parsers | **80.7%** ± 0.4% | 80.0% | 34s / 50s | $10.00 | $18.89 | 25.0M | [$0.010 / page](https://www.datalab.to/pricing) |
| 5 | [Reducto](https://docs.reducto.ai/) model=r-1 | Specialised document parsers | **84.8%** ± 0.3% | 85.9% | 5s / 12s | $10.00 | $17.97 | 26.4M | [$0.010 / page](https://reducto.ai/pricing) |
| 6 | [Datalab convert](https://documentation.datalab.to/) convert · mode=accurate | Specialised document parsers | **63.9%** ± 3.4% | 55.7% | 24s / 55s | $10.00 | $23.86 | 30.2M | [$0.010 / page](https://www.datalab.to/pricing) |
| 7 | [Mistral OCR](https://docs.mistral.ai/capabilities/OCR/basic%5Focr/) model=mistral-ocr-4-1 | Specialised document parsers | **36.9%** ± 0.4% | 16.8% | 15s / 72s | $4.00 | $16.54 | 51.2M | [$0.004 / page](https://mistral.ai/pricing/api/) |
| 8 | [Unstructured](https://docs.unstructured.io/transform/api/parseRun) parse · profile=best | Specialised document parsers | **27.0%** ± 0.2% | 2.6% | 65s / 119s | $15.00 | $84.67 | 70.3M | [$0.015 / page](https://docs.unstructured.io/transform/billing) |
| 9 | [Pulse](https://docs.runpulse.com/) model=pulse-ultra-2 | Specialised document parsers | **48.4%** ± 0.4% | 33.5% | 53s / 102s | $15.00 | $47.17 | 89.4M | [$0.015 / page](https://www.runpulse.com/pricing) |

\[02\] how it was measured

## How document-processing accuracy, latency, and cost were measured

This is the compact protocol. Corpus construction, all nine task types, parser settings, scoring rules, and cost treatment are documented in the [full benchmark methodology →](https://openbenchmarks.com/document-processing#methodology) 

* **Same corpus.** 94 legal contracts and 1,500 questions were reused for every measured configuration; no vendor received an easier document set.
* **One changing layer.** Every arm received the complete PDF and returned Markdown. The downstream reader, question, prompt, and scoring path stayed fixed, so the parser output was the variable under test.
* **Independent answer key.** Correct and stale answers were derived from tracked changes in the source Word files before vendor output was produced. The residual-answer judge was not shown the parser identity.
* **Bounded accuracy.** The best-case reference preserves every deletion; the markup-blind baseline removes every mark. Gap closed reports where each parser landed between those controls.
* **Separate operational metrics.** Parse latency and measured parser spend are reported beside semantic accuracy, never blended into a synthetic score. Published list price remains separate from measured corpus spend.
* **Reproducible evidence.** The benchmark runner and published data are open in [openbenchmarks-labs/document-processing ↗](https://github.com/openbenchmarks-labs/document-processing).

\[03\] how to use this result

## What this ranking establishes

### Semantic accuracy

A correct answer follows the language the parties agreed. A stale answer follows language they struck, while a fused answer combines deleted and surviving text into a value that never existed.

### Independent controls

The best-case reference preserves every deletion. The baseline removes every mark. Gap closed places each measured system between those controls instead of pretending 100% is always reachable.

### Scope

Document parsers create the text used by downstream extraction, search, RAG, and agent workflows. This benchmark tests whether that text preserves the operative meaning of heavily redlined contracts: every system received the same complete PDF, and a fixed reader answered the same questions from each parser's Markdown.

\[04\] faq

## Questions answered by this comparison

### Which specialised document parser is the most token-efficient?

Extend leads the specialised document parsers in this table at 22.5M tokens per 1k correct, 93.1% accuracy. The result comes from 1,500 questions across 94 contracts.

### What does document-parsing accuracy mean here?

Accuracy is the share of questions answered from the operative contract language. The markup-blind baseline scored 25.2%. Stale rate separately counts answers taken from deleted language.

### How should latency and cost be read beside accuracy?

Extend recorded 17s median parse time and $20.00 in measured parse cost per 1,000 pages. The full board also divides parse cost and the answering agent's tokens by correct answers, which shows what a parser costs for every 1,000 right answers it leads to.

### How was this document-processing comparison independently benchmarked?

Each system converted the same full PDFs to Markdown. The same reader model then answered the same hidden-ground-truth questions; vendor identity was not shown to the residual-answer judge.

### What documents does this benchmark cover?

Document parsers create the text used by downstream extraction, search, RAG, and agent workflows. This benchmark tests whether that text preserves the operative meaning of heavily redlined contracts: every system received the same complete PDF, and a fixed reader answered the same questions from each parser's Markdown.

\[05\] related comparisons

## Related document-processing pages

* [Document Processing Benchmark: complete leaderboard and methodology →](https://openbenchmarks.com/document-processing)
* [Best Document Parsing APIs](https://openbenchmarks.com/document-processing/best-document-parsing-apis)
* [Best OCR APIs for Scanned Documents](https://openbenchmarks.com/document-processing/best-ocr-apis-for-scanned-documents)
* [Best Document Parsing APIs by Cost](https://openbenchmarks.com/document-processing/best-document-parsing-apis-by-cost)

\[06\] source benchmark

## Read the complete Document Processing Benchmark

Task design, corpus transformation, parser settings, scoring, controls, cost accounting, and both tagged-PDF and scanned-PDF leaderboards live on the [Document Processing Benchmark →](https://openbenchmarks.com/document-processing#methodology) 

Openbenchmarks [document processing benchmark](https://openbenchmarks.com/document-processing) [all benchmarks](https://openbenchmarks.com/) © 2026 Openbenchmarks