OpenThinker AI Model Surpasses DeepSeek in Key Benchmarks
- OpenThinker-32B achieved a 90.6% accuracy on the MATH500 benchmark, surpassing DeepSeek’s 89.4%.
- The model scored 61.6 on the GPQA-Diamond benchmark, outperforming DeepSeek’s 57.6.
- It required only 114,000 training examples compared to DeepSeek’s 800,000.
- OpenThinker scored lower in coding tasks with a score of 68.9 versus DeepSeek’s 71.2.
The OpenThinker-32B model demonstrated superior performance over DeepSeek in specific benchmarks like MATH500 and GPQA-Diamond while using significantly fewer training examples, highlighting its efficiency and effectiveness.
Source (2.6)https://decrypt.co/305878/new-open-source-ai-model-rivals-deepseeks-performance-with-far-less-training-data?rand=52368
