Alibaba’s Qianwen released Qwen3.8-Flash, a multimodal Mixture-of-Experts (MoE) model and an early preview of the Qwen4 architecture. Production Qwen3.8-Flash will be available via the Qwen Cloud API, priced at $0.16 per 1M input tokens and $0.47 per 1M output tokens. The model includes 125 billion parameters plus 51 billion N-gram embedding parameters, with roughly 6 billion parameters activated per token to boost cost efficiency.

消息人士：巴西政府將延長對汽油的補貼至當地時間9月9日。

China International Capital Co says tantalum and niobium supply is highly fragile and increasingly exposed to external disruption. New demand from AI, new energy and commercial space sectors is tightening balances and, combined with geopolitical frictions, could lift price midpoints. The report flags strategic upside for fully integrated, full‑chain players — firms with secured upstream resources, expanding midstream capacity and defensible downstream technologies are likely to benefit materially.