GLM 5.2: what developers actually say
We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on GLM 5.2. 110 of the 142 opinions come from people who say they use it themselves.
Developers give GLM 5.2 a 75% approval score across 142 opinions from 109 authors, and 110 of those opinions come from firsthand use. Code quality is the strongest point at 41 positive mentions against 6 negative, with cost next at 30 against 5. Approval over the last 90 days is 82%, up from 72% in June 2026, but opinion volume fell from 59 in July to 3 in September and 4 in October, so the recent gain rests on thin data. Reliability is the weak spot at 7 positive against 9 negative, and some users report token burn that cancels out the low price.
Pick it if
- Want strong code quality at a low price
- Run long agent jobs on a small budget
- Want generous usage limits and an open model
Look elsewhere if
- Need steady output quality across runs
- Need vision or strong prompt adherence
- Need frontier results on the hardest tasks
Strengths and weaknesses
Each row counts opinions that mentioned the topic. Bars to the right of the line are praise, bars to the left are complaints.
How opinion has moved
What developers say
Praise
Complaints
Summaries are paraphrased. Each links to the original comment.
GLM 5.2 compared
FAQ
Is GLM 5.2 worth it?
Most users say yes. Approval is 75% overall and 82% over the last 90 days. The common case is cheap, capable code generation. Complaints center on uneven output and token costs that one user found higher than GPT-5.6.
Is GLM 5.2 better than Claude Opus?
The data is split. Two users preferred Claude Opus 4.8, while two preferred GLM 5.2 over Claude Code. One SVG test found GLM 5.2 behind Opus on output structure.
Is GLM 5.2 good for coding?
Yes for most work. Code quality gets 41 positive mentions against 6 negative. Users still report errors on hard tasks, weaker prompt adherence than Claude and GPT, and one account of 20-30% mistakes at Max reasoning.