Comment by ckocagil

Because it's notoriously hard to benchmark LLMs. Ultimately every benchmark is different and measures different things. This is why companies that make LLM models have private benchmarks - they find the areas where the model is weak and make that their goal.