The recent news revolves around Google's Gemini-1.5 Pro achieving top performance in a new OCR benchmark. However, the validity of this comparison has come under scrutiny. Critics argue the testing lacks fairness since the systems employed often reside within larger frameworks. They suggest using alternative benchmarks like MinerU, OHR Bench, and Reducto, which could yield a more accurate reflection of each model's capabilities. Furthermore, the comparison has sparked debates over the models' architectures and optimization goals. Observations indicate that while Gemini excels in metrics like word error rate, it's positioned against models that are considerably older and not strictly optimized for OCR tasks, like EasyOCR and RapidOCR. Some commentators have pointed out that results may also be influenced by the 'temperature' setting in transformer models which directly affects output creativity and accuracy. Users express dissatisfaction with current AI tools' complexity when handling various document formats for querying information effectively, highlighting an ongoing pain point in user experience and model practicality.