Recent releases from Chinese labs, including Alibaba’s Qwen3.8-Max and early Qwen4-Max preview with mixture-of-experts (MoE) scaling up to trillions of parameters, Moonshot’s Kimi K3 series, and DeepSeek’s frequent V4.1 Flash updates, have strengthened open-weight large language model (LLM) performance on coding and efficiency benchmarks. These models lead open-source leaderboards and offer dramatically lower API costs than Western counterparts, boosting developer adoption. However, composite indices from Artificial Analysis and Arena place the strongest Chinese entries around sixth to ninth globally, trailing closed US frontier systems by substantial margins on reasoning and multimodal tasks. With December 31 approaching, any additional 2026 releases or benchmark gains could narrow the gap, though historical patterns suggest the frontier divide may persist absent major breakthroughs.
Tóm tắt AI thử nghiệm tham chiếu dữ liệu Polymarket. Đây không phải tư vấn giao dịch và không ảnh hưởng đến cách thị trường này được giải quyết. · Cập nhật1H6H1D7D30DALL
View resolved

Cẩn thận với liên kết bên ngoài.
Cẩn thận với liên kết bên ngoài.
Câu hỏi thường gặp