Scale AI has introduced a new private leaderboard for evaluating large language models (LLMs), marking a significant development in the field of AI. This leaderboard aims to provide a more accurate and trusted evaluation of public models by using clean, third-party assessments. The initiative addresses concerns about overfitting and contamination in existing benchmarks like MMLU. Additionally, the release of K2-65B, a fully open-source LLM with 65 billion parameters, has been announced. K2-65B is noted for its transparency, reproducibility, and superior performance compared to Llama 2 70B. The model includes all necessary components for open-source AGI, such as model checkpoints, code, logs, and data, and is released under the Apache 2.0 license. LLM360 and SnowflakeDB have been active in supporting this open-source AI initiative, and the LMSys Arena is also seen as a complementary benchmark.