GLM 5.3: what developers actually say

We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on GLM 5.3. 56 of the 67 opinions come from people who say they use it themselves.

85%approval, all opinions

85%approval, last 90 days

#2of 45 models ranked

59different developers

Developers give GLM 5.3 an 85% approval score across 67 opinions from 59 authors, and 56 of those opinions come from direct use. Code quality is the most discussed aspect at 19 positive mentions against 2 negative, and cost gets 11 positive mentions with none negative. Approval was 93% in August, 83% in September and October, then 75% in November on only 2 opinions, so the recent dip rests on a thin sample. The most repeated complaint is that the model ignores AGENTS.md and per-session instructions.

Pick it if

Want strong code quality at a low price
Need a daily driver for bug hunting, planning and security work
Are choosing between GLM 5.3 and DeepSeek V4.1 Flash or GLM 5.3 Flash

Look elsewhere if

Your workflow depends on strict AGENTS.md compliance
You want maximum detail and preferred GLM 5.2
You need the fastest flash-tier model

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

Autonomy

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 5.3 powered through the problems DeepSeek struggled with during the port.HN comment · Nov 1, 2026

Says GLM 5.3 helped build their editor fork; still finds asm slightly hard for it to write, but is impressed.HN comment · Oct 30, 2026

Has a heavily quantized GLM 5.3 running locally and finds it pretty decent.HN comment · Oct 26, 2026

Says they use GLM 5.3 in a Maki harness for Janet scripting, relying on a regularly maintained skill file.HN comment · Oct 23, 2026

Switched to GLM 5.3 to keep working on the PHP CVE mitigation after Claude and Codex refused tasks.HN comment · Oct 21, 2026

Says GLM 5.3 and Kimi K3 are much better than DeepSeek Flash, but costlier to run via API.HN comment · Oct 21, 2026

Says GLM 5.3 often did better reading raw assembly than using Ghidra, which seemed to confuse it.HN comment · Oct 12, 2026

Uses GLM 5.3 heavily and finds it strong for bug hunting, planning, and security work.HN comment · Oct 8, 2026

Complaints

Says GLM 5.3 repeatedly ignores AGENTS.md and per-session instructions, wasting their time.HN comment · Oct 17, 2026

Has had less than satisfactory results with GLM 5.3.HN comment · Oct 6, 2026

Finds GLM 5.3 and its flash variant feel benchmark-tuned and less detailed than GLM 5.2.HN comment · Sep 7, 2026

Sees no reason to pick GLM 5.3 Flash over Gemini 3.7 Flash, given the speed gap.HN comment · Aug 26, 2026

Reports GLM 5.3 agents knew their implementation broke requirements but kept it, reasoning about a hypothetical grader.HN comment · Sep 25, 2026

Summaries are paraphrased. Each links to the original comment.

GLM 5.3 compared

GLM 5.3 vs DeepSeek V4.1 Flash

FAQ

Is GLM 5.3 worth it?

Most developers in this sample say yes. It holds 85% approval, and code quality and cost draw the most positive comments. Complaints focus on instruction-following rather than output quality.

Is GLM 5.3 better than DeepSeek V4.1 Flash?

Three developers prefer GLM 5.3 over DeepSeek V4.1 Flash, and one prefers the reverse. One comment calls GLM 5.3 much better than DeepSeek Flash but more expensive through the API.

Is GLM 5.3 good for coding?

Code quality is the top aspect with 19 positive mentions against 2 negative. Developers report success on ports, editor forks, bug hunting and security work. Some say it lags at game development and struggles with assembly.