v0.14.0
datalab-to/suryav0.14.0May 13, 2025by VikParuchuri
AI Summary
Major architectural overhaul introducing Surya OCR 3, trained on significantly more data. This release focuses on performance, expanding language support, and improving handling of complex layouts like inline math.
Key Highlights
- Support for 90+ languages with significantly better Chinese performance.
- Handles inline math and equations with char, word, and line bounding boxes.
- High speed performance: 10000+ tokens/second on A100 with vllm.
- Continuous batching in base configuration for approximately 2x speedup.
- New architecture and foundation processor.
New Features
- New foundation processor
- Continuous batching optimizations
- Qwen encoder integration
- Surya OCR 3 architecture
Full Release Notes
# Surya OCR version 3! The latest version of Surya OCR has a new architecture, and is trained on significantly more data than before. Some notable features: - 90+ language support - Handles inline math and equations - Char, word, and line bboxes available - Significantly better chinese performance - Very fast - 10000+ tokens/second on A100 with vllm - Continuous batching in base configuration for ~2x speedup Updated benchmarks coming soon. ## Examples Word boxes  Math <img width="1842" alt="image" src="https://github.com/user-attachments/assets/0b768cdf-74da-4482-99c8-2794e7964be3" /> Chinese <img width="1854" alt="image" src="https://github.com/user-attachments/assets/d6a34ad3-0707-4f65-923c-17917ee8002d" /> ## What's Changed * Foundation new processor by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/342 * WIP: Continuous Batching + Other Optimizations for new model by @tarun-menta in https://github.com/VikParuchuri/surya/pull/337 * Foundation vik dev by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/343 * Foundation vik dev by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/348 * Foundation vik dev by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/349 * Foundation vik dev by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/351 * Vik perf by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/353 * Vik perf by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/355 * Foundation flash by @tarun-menta in https://github.com/VikParuchuri/surya/pull/354 * Vik perf by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/356 * Vik perf by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/358 * Qwen encoder by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/362 * Foundation by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/363 * Vik dev by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/365 * Surya OCR 3 by @VikParuchuri in https://github.com/VikParuchuri/surya/pull/364 **Full Changelog**: https://github.com/VikParuchuri/surya/compare/v0.13.1...v0.14.0