According to CoinMeta, the latest risk report from Anthropic reveals for the first time the internal model “Model 2”, which performs better overall than “Mythos 5” and has seen significant improvements in multiple internal tasks. Currently, this model is widely used for writing code, generating data, and running “agent”, but there are no plans to release “Anthropic” to the public for now. In addition, the report indicates that the model’s risk assessment in high-risk scenarios has been raised from “extremely low” to “low” due to recent security testing incidents, which have made the company less confident in its assessment of such risks. The report also mentions that an internal model codenamed “Model 2” has performed exceptionally well in many tasks, suggesting that the “Anthropic” model, which is perceived by the public as the strongest, may actually be behind the company’s true cutting-edge models.