Models
OpenTyphoon.ai offers several models optimized for Thai language understanding and generation. Each model has different capabilities, performance characteristics, and rate limits.
Available Models
Section titled “Available Models”| Model ID | Size | Description | Context Window | Rate Limits | Release Date |
|---|---|---|---|---|---|
| typhoon-v2.1-12b-instruct | 12B | Our most powerful model with advanced Thai language capabilities | 56K tokens | 5 req/s, 200 req/m | 2025-05-05 |
| typhoon-ocr | - | Specialized OCR & document parsing model | - | 2 req/s, 20 req/m | 2025-05-19 |
Model Details
Section titled “Model Details”typhoon-v2.1-12b-instruct
Section titled “typhoon-v2.1-12b-instruct”Our previous flagship model, typhoon-v2.1-12b-instruct, is designed for advanced Thai language understanding and generation. It excels at:
- Complex Thai language understanding and generation
- Following detailed instructions in both Thai and English
- Maintaining context over longer conversations
- Code understanding and generation
- Creative writing and content generation in Thai
Best For: Applications requiring the highest quality Thai language understanding and generation, including customer support, content creation, and complex question answering.
typhoon-ocr
Section titled “typhoon-ocr”Our specialized OCR & document parsing model, typhoon-ocr is designed specifically for:
- Optical character recognition optimized for Thai text
- Document parsing and information extraction
- Structured data extraction from images and documents
- Form processing and data capture
Best For: Applications that need to extract text and data from images, scanned documents, forms, and other visual media containing Thai text.
Choosing the Right Model
Section titled “Choosing the Right Model”Consider these factors when choosing a model:
- Task Complexity: More complex tasks generally benefit from larger models
- Response Speed: Smaller models generally have lower latency
- Rate Limits: Consider your application’s volume requirements
- Quality Requirements: Higher quality generally requires larger models
For most applications requiring high-quality Thai language understanding and generation, we recommend using typhoon-v2.1-12b-instruct. For OCR and document parsing needs, use typhoon-ocr.
Research and Development
Section titled “Research and Development”Our models are built on extensive research in Thai language processing. You can learn more about the technical details in our research papers:
Future Models
Section titled “Future Models”We’re continuously working on improving our models and will release new versions as they become available. Stay updated by following our blog or joining our Discord community.