GLM 4.7: what developers actually say

We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on GLM 4.7. 39 of the 43 opinions come from people who say they use it themselves.

56%approval, all opinions

–approval, last 90 days

#38of 45 models ranked

38different developers

Developers give GLM 4.7 a 56% approval score across 43 opinions from 38 people, and 39 of those opinions come from firsthand use. Cost is the clear strength: 9 of 10 pricing comments are positive, and users cite cheap coding plans with generous limits. Code quality is the weak spot at 7 positive and 10 negative comments, and reliability drew 3 complaints with none positive. Approval has drifted down from 62% in January 2026, when 24 opinions landed, to 38% in March and 33% in June. Those later months hold only 1 to 4 opinions each.

Pick it if

Want a cheap coding plan with generous usage limits
Need a pair programmer for routine debugging and small scripts
Are cutting costs on a multi-agent Claude Code setup

Look elsewhere if

Work on large C++ or Rust changes where intent must hold
Need output you can trust without extra fixing
Expect problem solving close to Sonnet or Opus

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.

PraiseComplaints

Code quality

Cost & pricing

Reliability

Speed

How opinion has moved

Monthly approval (line) and number of opinions (bars). Months with fewer than 3 opinions are left out.

What developers say

Praise

Says GLM 4.7 on Cerebras used to be amazing and wants much faster inference.HN comment · Aug 14, 2026

Says GLM 4.7 was good enough to help with a lot of work when used as a pair programmer.HN comment · Jun 22, 2026

Says GLM 4.7 on Cerebras is their current pick: the smallest, fastest, cheapest model that gets the job done.HN comment · Feb 12, 2026

Says GLM 4.7 has been performing very well so far.HN comment · Feb 4, 2026

Uses GLM 4.7 via a cheap coding plan that gives about 3x Claude Pro's usage.HN comment · Feb 2, 2026

Says GLM 4.7 is surprisingly good at debugging complex Neovim configs.HN comment · Jan 29, 2026

Says GLM 4.7 gave promising results, though not yet matching the paid models.HN comment · Jan 28, 2026

Uses GLM 4.7 to cut costs and reports no issues running a multi-agent Claude Code setup.HN comment · Jan 23, 2026

Complaints

Says GLM 4.7 Flash was far behind frontier models at launch, based on paid use and running it locally.HN comment · Jun 25, 2026

Says GLM 4.7 was extremely bad for agents writing simple 200-1000 line scripts.HN comment · Jun 8, 2026

Says the GLM 4.7 model they used on Cerebras is outdated.HN comment · May 20, 2026

Says GLM 4.7 clearly trails Sonnet and Opus in problem-solving ability.HN comment · Apr 4, 2026

Local GLM 4.7 gets stuck fixing compiler errors and loses the intent of changes in C++ and Rust.HN comment · Mar 27, 2026

Says GLM 4.7 needed more fixing of its output than they'd like.HN comment · Mar 7, 2026

GLM 4.7 guessed dataset details from a URL and built a broken app backed by fake-data tests.HN comment · Feb 12, 2026

Says GLM-4.7 used in Claude Code also drifts off track without a planning step.HN comment · Feb 2, 2026

Summaries are paraphrased. Each links to the original comment.

FAQ

Is GLM 4.7 worth it?

For cost-sensitive work, the data is favorable. Pricing is the strongest aspect at 9 positive and 1 negative comment. Code quality is the weakest at 7 positive and 10 negative, and reliability has 3 complaints and no praise.

Is GLM 4.7 better than Claude Opus?

Several users rank it below Sonnet and Opus. One describes it as less clever than Opus 4.5 while often getting the job done. One person preferred it over Claude Code.

Is GLM 4.7 good for coding?

The results split. Users report good results on debugging complex Neovim configs and on small pair-programming tasks. Others report failures on agent-written scripts of 200 to 1000 lines and uncertain quality on a C++ to Rust port.