Hugging Face OpenThaiGPT R1 32b: Advanced Thai Reasoning Model
OpenThaiGPT R1 32b is a 32-billion-parameter model specifically designed for advanced Thai language reasoning. Developed by the OpenThaiGPT Team, this model demonstrates superior performance in complex analytical tasks compared to larger 70B models like DeepSeek R1 70b and Typhoon R1 70b, despite being less than half their size. It is particularly strong in mathematical, logical, and code reasoning, offering explicit step-by-step thought processes.
Key Capabilities
- State-of-the-art Thai Reasoning: Outperforms larger models on mathematical and logical reasoning benchmarks in Thai.
- Explicit Reasoning: Capable of showing detailed, step-by-step thought processes for problem-solving.
- Efficient Performance: Achieves high reasoning capabilities with a significantly smaller parameter count (32B) compared to 70B alternatives.
- Code Reasoning: Excels in code reasoning tasks in both Thai and English.
Benchmark Highlights
OpenThaiGPT R1 32b leads across several benchmarks:
- AIME24-TH: 56.67 (vs. DeepSeek R1 70b: 33.33, Typhoon R1 Distill 70b: 53.33)
- AIME24: 63.36 (vs. DeepSeek R1 70b: 53.33, Typhoon R1 Distill 70b: 53.33)
- MATH500-TH: 83.8 (vs. DeepSeek R1 70b: 75.4, Typhoon R1 Distill 70b: 81)
- LiveCodeBench-TH: 62.16 (vs. DeepSeek R1 70b: 53.15, Typhoon R1 Distill 70b: 47.75)
- LiveCodeBench: 69.67 (vs. DeepSeek R1 70b: 64.97, Typhoon R1 Distill 70b: 54.79)
Its average performance across these benchmarks is 71.58, surpassing both DeepSeek R1 70b (63.31) and Typhoon R1 Distill 70b (65.42).
Use Cases
This model is ideal for applications requiring robust Thai language understanding and complex problem-solving, especially in educational technology, automated code analysis, and advanced logical reasoning systems. It is available for both research and commercial use.