QWEN 2.5-Max exceeds Deepseek V3 with some benchmarks

Alibaba’s reaction to DeepSeek is Qwen 2.5-Max, a large-scale model of the company’s latest EXPERTS (MOE). QWEN 2.5-Max has deleted more than 20 trillion tokens in advance, and boasts fine-tuning through state-of-the-art techniques such as monitored fine-tuned (SFT) and reinforcement learning from human feedback (RLHF). The API is available via Alibaba Cloud and can access [...]

The post QWEN 2.5-Max exceeds Deepseek V3 with some benchmarks first appeared on Versa AI hub.

from Blog - Versa AI hub https://versaaihub.com/qwen-2-5-max-exceeds-deepseek-v3-with-some-benchmarks/
via IFTTT