Yes very interesting potential, it looks like it can be increased in accuracy considerably because Llama 3.1 with 405B parameters has very similar performance with the latest GPT-4o.