openthaigpt/openthaigpt-1.6-72b-instruct

TEXT GENERATIONPricing:Input $12 / Output $20Concurrent Unit Cost:4Model Size:72.7BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Mar 25, 2025License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold

OpenThaiGPT 1.6 72b is a 72-billion-parameter Thai language model developed by OpenThaiGPT, designed for general-purpose tasks with strong reasoning capabilities. This instruction-tuned model demonstrates improved performance in code reasoning and general language tasks compared to previous versions. It offers balanced capabilities across mathematical, coding, and general language understanding, making it suitable for a wide range of applications requiring advanced Thai language processing.

Loading preview...

OpenThaiGPT 1.6 72b: Advanced Thai Language Model

OpenThaiGPT 1.6 72b is a 72-billion-parameter model from OpenThaiGPT, building upon the OpenThaiGPT 1.5 foundation. It is specifically designed for general-purpose tasks in Thai, featuring enhanced reasoning capabilities in both Thai and English. The project, now known as OpenThai, continues to develop models tailored for the Thai language and cultural context.

Key Capabilities & Improvements

  • Advanced Thai Language Processing: A 72-billion-parameter model with enhanced understanding of Thai language and cultural nuances.
  • Strong Reasoning: Demonstrates robust reasoning abilities across mathematical, coding, and general language tasks.
  • Improved Performance: Shows significant improvements over previous OpenThaiGPT versions, particularly in code reasoning.
  • Benchmark Highlights: Achieves 32.43 on LiveCodeBench-TH and 54.21 on LiveCodeBench, indicating strong coding performance. It also scores 82 on MATH500 and 78.7 on OpenThaiEval.
  • Balanced Functionality: Offers a good balance between general language understanding and specialized capabilities like coding and mathematics.

When to Use This Model

  • General Thai Chat and Coding: Ideal for applications requiring comprehensive Thai language interaction and code generation.
  • Reasoning Tasks: Suitable for use cases demanding strong logical and mathematical reasoning.
  • Research and Commercial Applications: Available for both research and commercial use under specified license terms.
  • Local Deployment: Can be run locally using Ollama or vLLM, with GGUF quants available for memory-constrained environments (e.g., ~48 GB RAM/VRAM for 4-bit quantization).

Finetunes
2 models