Alibaba Cloud launches four open-source models: Qwen3.5, GLM-5, MiniMax M2.5, and Kimi K2.5

robot
Abstract generation in progress

On February 25th, Alibaba Cloud’s BaiLian launched the Coding Plan, which includes API services for four major open-source models: Qwen3.5, GLM-5, MiniMax M2.5, and Kimi K2.5. After subscribing to a plan, users are no longer limited to a single model and can freely switch between multiple models, enjoying more stable and higher token quota model services.

Currently, Alibaba Cloud’s Coding Plan has launched eight programming models, with the four newly released models all being excellent open-source models. The native multimodal model Qwen3.5, open-sourced on New Year’s Eve, outperformed expectations, setting new standards for efficiency and cost-effectiveness in large models. The world’s largest AI open-source community, Hugging Face, announced the latest open-source large model rankings, with Qwen3.5 topping the list. GLM-5 is the latest flagship open-source model from Zhipu, and its API service is in high demand. MiniMax M2.5 is a model specifically designed for Agent scenarios, performing well in advanced Excel processing, in-depth research, PowerPoint, and other Office tasks. Kimi K2.5 has also achieved significant improvements in programming and intelligent agent capabilities.

According to reports, Alibaba Cloud’s Coding Plan Lite basic package offers up to 18,000 requests per month, with new users paying only 7.9 yuan for the first month. The Pro advanced package provides 90,000 requests per month, capable of handling thousands of complex tasks, with the first month costing only 39.9 yuan, greatly reducing costs for high-frequency coding scenarios. After subscribing, users can seamlessly switch between models such as Qwen Code, Claude Code, Cline, and OpenClaw across various AI tools.

The enhancements in programming and agent capabilities are expected to promote more diverse AI applications. Some industry analysts predict that China’s AI agent market will surpass 3.3 trillion yuan by 2028. However, because agents need to perform complex tasks involving multi-round planning and tool invocation, token consumption increases exponentially. The industry still faces the challenge of addressing the rising costs associated with agent applications. Data shows that the computational power required for a single agent can be 100 to 1,000 times that of a traditional chatbot.

View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
0/400
No comments
  • Pin