AI Handwriting Recognition Rankings
As genealogists, using artificial intelligence to transcribe handwritten historical documents is one of the most valuable applications of the technology. Our free AI Handwritten Text Recognition Tool can help you batch-transcribe documents using many different providers through an easy-to-use interface, but choosing the right model can be tricky. The rankings below highlight which current models deliver the best performance, speed, price and value for your family history transcription project.
How to read these results
- Score: Word-level accuracy percentage averaged across multiple historical documents.
- Speed: Average time in seconds the model takes to complete the transcription of one image.
- $/100: The average total cost in USD to transcribe 100 images using this model via OpenRouter.
- Value: A score from 1-10 considering accuracy and cost efficiency (higher is a better value).
| Model Name | Score | Speed | $/100 | Value |
|---|---|---|---|---|
| amazon/nova-lite-v1 | 84.16% | 5.18s | $0.03 | 8.88 |
| anthropic/claude-3.5-haiku | 84.62% | 8.76s | $0.34 | 8.73 |
| anthropic/claude-haiku-4.5 | 78.57% | 6.23s | $0.43 | 8.26 |
| anthropic/claude-sonnet-4 | 91.50% | 11.61s | $1.30 | 8.68 |
| anthropic/claude-3.7-sonnet | 94.99% | 8.33s | $1.30 | 8.93 |
| anthropic/claude-sonnet-4.5 | 90.32% | 11.55s | $1.31 | 8.60 |
| google/gemini-2.0-flash-001 | 95.60% | 4.71s | $0.06 | 9.66 |
| google/gemini-2.0-flash-lite-001 | 95.14% | 4.13s | $0.03 | 9.64 |
| google/gemini-2.5-flash | 95.67% | 4.47s | $0.18 | 9.60 |
| google/gemini-2.5-flash-lite | 92.52% | 3.59s | $0.04 | 9.45 |
| google/gemini-2.5-flash-lite-preview-09-2025 | 83.02% | 2.01s | $0.02 | 8.80 |
| google/gemini-2.5-flash-preview-09-2025 | 67.10% | 5.25s | $0.13 | 7.62 |
| google/gemini-2.5-pro | 96.45% | 18.45s | $3.08 | 8.04 |
| google/gemma-3-12b-it | 40.11% | 5.38s | $0.01 | 5.80 |
| google/gemma-3-27b-it | 45.28% | 7.31s | $0.02 | 6.17 |
| meta-llama/llama-4-maverick | 82.12% | 14.54s | $0.09 | 8.70 |
| meta-llama/llama-4-scout | 79.95% | 5.87s | $0.05 | 8.57 |
| mistralai/mistral-small-3.2-24b-instruct | 89.75% | 8.22s | $0.04 | 9.26 |
| openai/gpt-4.1 | 92.43% | 11.39s | $0.60 | 9.14 |
| openai/gpt-4.1-mini | 95.71% | 7.77s | $0.20 | 9.59 |
| openai/gpt-4.1-nano | 81.71% | 6.68s | $0.07 | 8.68 |
| openai/gpt-4o | FAILED | - | - | - |
| openai/gpt-4o-mini | FAILED | - | - | - |
| openai/gpt-5 | 92.45% | 108.87s | $5.40 | 6.47 |
| openai/gpt-5-mini | 96.10% | 33.75s | $0.58 | 9.40 |
| openai/gpt-5-nano | 49.55% | 53.33s | $0.24 | 6.34 |
| openai/o3 | 94.57% | 43.40s | $2.82 | 8.05 |
| openai/o4-mini | 95.59% | 43.53s | $1.86 | 8.66 |
| openai/o4-mini-high | 74.67% | 59.59s | $3.84 | 6.09 |
| openrouter/polaris-alpha | 90.90% | 7.91s | $0.00 | 9.36 |
| qwen/qwen3-vl-235b-a22b-instruct | 96.80% | 9.07s | $0.14 | 9.70 |
| x-ai/grok-4 | 33.04% | 90.32s | $6.82 | 2.31 |
| x-ai/grok-4-fast | 22.99% | 10.60s | $0.06 | 4.58 |