This NEW Chinese AI is INSANE! (FREE + Open Source!)

Long Cap 2.0 is a newly released Chinese AI model that has generated significant attention despite its unconventional origins. Rather than coming from a major technology corporation, the model was developed by a food delivery app company. The model initially launched anonymously on OpenRouter under the code name Alpha, where it operated for approximately two months before gaining recognition and climbing performance charts without its creator being publicly identified. Long Cap 2.0 features 1.6 trillion parameters in total, but employs a sophisticated selective activation mechanism that only activates the parameters necessary for each specific task. This approach provides substantial computational benefits by reducing power consumption and increasing processing speed compared to models that activate all parameters uniformly. The most distinctive feature is its context window capacity of 1 million tokens, allowing users to input and process entire knowledge bases simultaneously while maintaining all information in active memory. In terms of performance benchmarks, Long Cap 2.0 achieved a score of 59.5 on SWE bench pro, exceeding GPT-4.5's score of 58.6. The model is released under the MIT open-source license, making it freely available for anyone to use, modify, and build upon commercially or otherwise.