Alibaba: qwen-mt-turbo API

Qwen-mt-turbo is the high-efficiency, speed-optimized machine translation model within the Qwen family. Built on a lightweight Qwen3-based Mixture-of-Experts (MoE) or distilled dense architecture, it is engineered for low-latency, high-concurrency environments such as live chat, social media localization, and real-time news feeds. Supporting 92 languages, it achieves a superior balance between speed and translation fidelity, often outperforming much larger general-purpose models in specific linguistic fluency benchmarks. While it provides "Good" quality compared to the "Best" quality of the Plus version, it significantly reduces the time-to-first-token, making it ideal for mobile and web applications that require near-instant cross-lingual communication.

• Context window: 16,384 tokens
• Max output: 8,192 tokens
• Input: text
• Output: text
• Knowledge cutoff: 2025-10-31