The AI arms race continues

There, Alibaba Qwen introduced a new AI model called Qwen2.5-Max.

The team claimed that it outperforms solutions such as GPT-4o, DeepSeek-V3, Llama-3.1-405B, and Claude 3.5 Sonnet.

According to the official release, the model was trained on 20+ trillion tokens using SFT and RLHF. Performance was evaluated using benchmarks MMLU-Pro, LiveCodeBench, LiveBench, and Arena-Hard.

This time, the stock and cryptocurrency markets did not react to this news, unlike with DeepSeek. NVIDIA shares also remained at the same level — apparently, the market has already digested this whole story.