Qwen 3 Coder: what developers actually say

We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on Qwen 3 Coder. 27 of the 39 opinions come from people who say they use it themselves.

87%approval, all opinions

–approval, last 90 days

#3of 45 models ranked

35different developers

Developers give Qwen 3 Coder an 87% approval across 39 opinions from 35 authors, and 27 of those opinions are firsthand. Code quality carries the score at 17 positive mentions against 3 negative, with speed at 7 positive and 0 negative and local privacy at 6 positive and 0 negative. Cost draws 3 positive mentions and no complaints, and users report it useful for boilerplate and autocomplete on older hardware. Monthly approval held at 100% from February to July 2026, then fell to 0% on 2 opinions in October 2026.

Pick it if

Want a local model for boilerplate and autocomplete
Care about privacy and running on your own hardware
Want solid code quality at low or zero cost

Look elsewhere if

Need dependable tool calls when driving MCP tools
Expect frontier-level quality from the 30B local build
Plan to pick the 480B model over GLM 4.6

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

Speed

Privacy & local use

Reliability

Cost & pricing

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 Qwen3-Coder 30B is more responsive and accurate for writing code than Qwen 3.6 in their tests.HN comment · Jun 30, 2026

Says Qwen 3 Coder clearly beats Qwen 3.6 for writing code, based on their own use on a server.HN comment · Jun 30, 2026

Runs Qwen3-coder:30b on an old GTX 1070 machine at zero cost and finds it useful for boilerplate.HN comment · Jun 15, 2026

Switched back to Qwen3-Coder on the older rig, preferring it over Qwen3.6.HN comment · Jun 15, 2026

Says tool-calling accuracy has been solid with local Qwen3 Coder models for months.HN comment · May 20, 2026

Says Qwen3-Coder-Next is still their favorite local model for coding.HN comment · Apr 30, 2026

Says Qwen3-Coder wrote much better Rust than Claude Opus or Gemini a few months ago.HN comment · Apr 20, 2026

Stopped using agentic AI; local Qwen3 Coder autocomplete is pretty fantastic and fast.HN comment · Apr 16, 2026

Complaints

Says Qwen3-Coder 30B hallucinates tool calls when driving an MCP tool.HN comment · Oct 2, 2026

Calls Qwen3 Coder 30B underwhelming compared with frontier models for local use.HN comment · Jan 23, 2026

Quotes research finding Qwen3-Coder's code about 130% more vulnerable for a government customer.HN comment · Oct 26, 2026

Says the 480B Qwen 3 Coder variant is largely worse than the 355B GLM 4.6 model.HN comment · Nov 4, 2025

Summaries are paraphrased. Each links to the original comment.

FAQ

Is Qwen 3 Coder worth it?

Approval sits at 87% across 39 opinions, and 27 come from people with direct experience. Code quality leads at 17 positive mentions against 3 negative. Users describe it as a capable local coding model, mainly for routine work.

Is Qwen 3 Coder better than GLM 4.6?

For the large models, the comments favor GLM 4.6. One user called the 480B Qwen 3 Coder worse than the 355B GLM 4.6, and another said it was well behind GLM 4.6 in quality. Qwen 3 Coder gets credit for price and speed in the same comparison.

Does Qwen 3 Coder work well for local use?

Privacy and local use get 6 positive mentions and no negatives. Users run the 30B build on an old GTX 1070 at zero cost and like it for boilerplate. One user finds it slow without an agentic harness, and another calls the 30B model underwhelming next to frontier models.