Gemma 4 and what makes an open model succeed

Open AI models are increasingly struggling to evaluate their performance, especially with the emergence of closed models like Claude and GPT. In recent years, open models have been scarce, making updates like Llama 3 highly anticipated. However, with rising competition from models such as Qwen 3.5 and Kimi K2.5, identifying the actual advantages of open models has become challenging. Open models require a different evaluation approach, as traditional criteria may not accurately reflect their true potential. Despite their significant capabilities, there is a lack of clear guidelines to harness these potentials.