Qwen2.5-VL-72B

Training compute

9.6×10²⁴ FLOP

Parameters

72B

Published

Feb 19, 2025

Qwen2.5-VL-72B is an AI model developed by Alibaba (China), first published in February 2025. It works in the language, vision and multimodal domain, on tasks such as visual question answering, video description, language modeling/generation and 4 more.

Training it took an estimated 9.6×10²⁴ FLOP of compute (estimation method: operation counting). The model has 72,000,000,000 parameters.

Access: Open weights (restricted use). Its weights are openly available. It is built on top of Qwen2.5-72B. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Alibaba
Country of organization: China
Domain: Language, Vision, Multimodal
Task: Visual question answering, Video description, Language modeling/generation, Translation, Question answering, Character recognition (OCR), Quantitative reasoning
Training compute: 9.6×10²⁴ FLOP
Compute estimation method: Operation counting
Parameters: 72,000,000,000
Model accessibility: Open weights (restricted use)
Open weights: Yes
Base model: Qwen2.5-72B
Epoch confidence: Confident