What is MacOCRBench?
MacOCRBench is an OCR benchmark designed for Apple Silicon Macs.
It measures OCR performance, throughput, multi-process efficiency, and stability, then evaluates the results as a Mac OCR PowerScore.
Requirements
- Apple Silicon Mac
- macOS Monterey or later
Download
MacOCRBench can be downloaded from:
Running the Benchmark

- Launch: Generate Test Images
- Automatically generates 20 OCR test images on your local machine.
- Start Test: Increase Concurrent Workloads
- When the benchmark starts, it runs a Single phase, followed by Multi 2-Lane, Multi 3-Lane, and additional phases with increasing workloads.
- Test Completion
- The benchmark automatically stops when a higher score can no longer be achieved.
- Results are provided as an HTML report.
HTML Report Example
Benchmark results are provided in the following format:
What is a Lane?
A Lane represents the number of OCR processes running concurrently.
Test Modes
| Test Mode | Description |
|---|---|
| Quick Test | Runs each phase once. Recommended for fanless systems such as MacBook Air models. |
| Standard Test | Recommended for general use. Each phase is executed three times, and the best score is recorded. |
| Research Test | Intended for detailed comparison and analysis. Each phase is executed five times, and the best score is recorded. |
Completion Criteria
The benchmark ends when a phase fails to exceed the score of the previous phase.
- MacOCRBench evaluates system performance based on the phase that achieves the highest score.
- Phases after performance saturation are skipped because further gains are unlikely.
Rankings and Phase Evaluation
After each phase completes, MacOCRBench evaluates rankings based on score and throughput, including:
- Ranking within the same Apple Silicon family
- Overall global ranking
Evaluation results are displayed both in the main application window and in the HTML report.
Tier
Tier provides an easy-to-understand overall performance classification.
| Tier | Criteria |
|---|---|
| S | Top 0–3% |
| A | Top 3–15% |
| B | Top 15–40% |
| C | Top 40–75% |
| D | Bottom 25% |
Rank
| Rank | Criteria |
|---|---|
| Elite | Top 0–5% |
| Excellent | Top 5–15% |
| Above Average | Top 15–30% |
| Average | Top 30–60% |
| Entry | Bottom 40% |
Scaling Efficiency
Scaling Efficiency measures how effectively performance improves as the number of lanes increases.
| Label | Criteria |
|---|---|
| Excellent | 90%+ |
| Good | 75%+ |
| Fair | 55%+ |
| Limited | 35%+ |
| Saturated | Below 35% |
NOTE
Benchmark results are submitted for ranking evaluation.
Submitted data:
- Score
- Model Name
- Memory Capacity
No personally identifiable information is collected.
An internet connection is required for ranking evaluation.
Reports
Benchmark results are provided as HTML files.
- File → Open Last Report...
- Opens the most recent benchmark report in your web browser.
- File → Open in Finder
- Opens the temporary report folder in Finder.
- File → Save Report...
- Saves the most recent benchmark report to a location of your choice.
Rankings
Ranking submission is optional.
Your results will only appear on the leaderboard if you choose to submit them.
Results can be published on both the Global Ranking and the family-specific ranking pages for:
- M1 Family
- M2 Family
- M3 Family
- M4 Family
- M5 Family
at the Ranking Site.
INFO
The ranking site will be published once a sufficient number of benchmark results have been collected.
Thank you for your patience.