Google has recently released Gemini 1.5, a model equipped with a context window of 1 million tokens, capable of processing entire books or even movies. However, during testing, the accuracy of Gemini 1.5 was found to be subpar, particularly in the "searching for a needle in a haystack" test, where its average accuracy ranged from 60% to 70%. Additionally, Google has questioned the authenticity of videos generated by OpenAI Sora, labeling them as fake.