Paid2026-02-23
09 To What Extent Has the Coding Ability of Domestic Models Progressed?
Recently, I tested two domestic models: GLM-5 and Qwen 3.5 Plus.
In 2025, domestic models couldn't really be used for coding, or at best, they were just "passable"—hand-holding them was very difficult.
As it turns out, the historical turning point occurred just a couple of days ago during the Spring Festival.
Over the past year, my requirements for a model's coding ability have also changed significantly. I have defined two stages:
- AI Programming 1.0 (2023 to before December 2025)
In this stage, an AI capable of writing code must meet two requirements:
- Able to understand requirements: comprehend what the development task is about.
- Able to write correct code: don't have frequent syntax errors; those are too low-level.
- AI Programming 2.0 (after December 2025 to who-knows-when; I'm not even sure what AI Programming 3.0 will look like)
In this stage, an AI with strong programming ability must not only meet the requirements of AI Programming 1.0 but also...
09 To What Extent Has the Coding Ability of Domestic Models Progressed? Recently, I tested two domestic models: GLM-5 and Qwen 3.5 Plus. In 2025, domestic models couldn't really be used for coding, or at best, they were just "passable"—hand-holding them was very difficult. As it turns out, the historical turning point occurred just a couple of days ago during the Spring Festival. Over the past year, my requirements for a model's coding ability have also changed significantly. I have defined two stages: 1. AI Programming 1.0 (2023–before December 2025). In this stage, an AI capable of writing code must meet two requirements: * Able to understand requirements:
Purchase required to continue
This is a paid article. After signing in, your purchase will be unlocked automatically.
Comments
No comments yet. Be the first to share your thoughts.
Leave a comment
No comments yet. Be the first to share your thoughts.