Google questions Gemini 4 performance

Google is facing an internal debate over the performance of its new AI model, Gemini 4, amid expectations to expand its launch in the face of strong competition from OpenAI and Anthropic. Despite achieving excellent results in standardized tests, reports indicate that the model struggles with real-world performance, especially in programming tasks, with some employees believing there is a gap between its test performance and its ability to handle actual tasks. Internal assessments vary regarding its level of advancement; some consider it worthy of being among the top-tier models, while others believe competitors outperform it in certain areas. Google has gradually released the Gemini 4 Argon model for complex tasks such as cybersecurity, implementing safety measures to mitigate risks. However, test results still do not fully reflect its practical performance, posing challenges to its quality and reliability in real-world applications.

Size: 120%Text Size

Notice: This Is an AI-Generated Summary

Display The Full Article