Alibaba

Traictory tracks 41 models from Alibaba. The earliest entry is Qwen2 72B Instruct from July 2024; the most recent is Qwen3.8 Flash, released in August 2026. 19 of them shipped in the last twelve months.

The strongest Alibaba model in our data is Qwen3 VL 32B Thinking, with an average benchmark score of 14450.8% across 6 published results, including 85500.0% on OCRBench and 830.0% on MM-MT-Bench. Across the 41 Alibaba models with benchmark data the average is 427.2%.

19 of the models are available through an API, with input pricing from $0.09 to $2.50 per million tokens; Qwen2.5-Coder 32B Instruct is the cheapest. The largest context window is 1M tokens on Qwen3.8 Flash Next. 18 of the 41 models accept images or other non-text input. 32 models ship with downloadable weights; the rest are API-only.