QWEN2

Alibaba Cloud's flagship open-source LLM series featuring 72B parameters and 128k context windows for high-tier multilingual reasoning.

Qwen2 delivers a powerful suite of large language models (outperforming Llama 3 70B in benchmarks like MMLU and HumanEval). The flagship 72B model utilizes a 128k context window and 27 trillion tokens of training data to master coding, mathematics, and 30 different languages. Five distinct model sizes (0.5B to 72B) provide the flexibility to deploy on local edge hardware or scale for heavy-duty enterprise automation.

https://github.com/QwenLM/Qwen2