Alibaba Qwen Releases Qwen3.8-Max: A 2.4 Trillion Parameter MoE Model and the Most Capable in the Qwen Family

Alibaba's Qwen team has made Qwen3.8-Max broadly available, with open weights promised next week. The 2.4-trillion-parameter mixture-of-experts model supports text, image, and video inputs, and outperforms its predecessor in multimodal and agentic tasks.

Alibaba/Qwen OpenAI Anthropic

DeepSeek-V4-Flash Official Benchmarks: About 98% Cache Hit for Ultimate Cost-Effectiveness

DeepSeek has released official benchmarks for its V4-Flash model, highlighting an approximately 98% cache hit rate. This is said to deliver extreme cost-performance, as cached tokens are significantly cheaper. The model has been presented as a competitive option in the AI inference market.

DeepSeek