featherless-ai/QRWKV-72B

featherless-ai/QRWKV-72B is a 72 billion parameter language model developed by featherless-ai, based on the RWKV architecture with a 65536 token context length. This model is a conversion of Qwen 2.5 72B into an RWKV variant, leveraging linear attention for significantly reduced computational costs at scale. It excels in general language understanding and generation tasks, demonstrating competitive performance against its Qwen2.5 counterpart across various benchmarks.