Qwen 3.8: what developers actually say

We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on Qwen 3.8. 47 of the 53 opinions come from people who say they use it themselves.

50%approval, all opinions

50%approval, last 90 days

#43of 45 models ranked

48different developers

Developers are split on Qwen 3.8. Approval is 50% across 53 opinions from 48 authors, and it is also 50% over the last 90 days. Code quality drew 8 positive and 7 negative comments, while reliability drew 3 positive and 6 negative. Monthly approval fell from 64% in August 2026 to 32% in October and 0% in November, though November has only 3 opinions.

Pick it if

Running a private local coding model on an M4 Mac or V100 GPUs
Handling basic CRUD work and config debugging on a budget
Batch jobs where a 27B model at Q5 or Q6 is enough

Look elsewhere if

Working in large-context projects that need to stay on track
Using low-spec hardware or 4-bit quantized builds
Needing fast iteration or frontier reliability for agentic coding

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

Reliability

Speed

Privacy & local use

Autonomy

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

After their own tests, says Qwen 3.8 is fine as a general-purpose model.HN comment · Oct 7, 2026

Says they use Qwen 3.8 on their coding team and rate it as good.HN comment · Oct 2, 2026

Says Qwen 3.8 is about as impressive as DeepSeek after trying it.HN comment · Sep 29, 2026

Runs Qwen 3.8 27B locally on an M4 Mac and calls it the first local model with consistently usable results.HN comment · Sep 26, 2026

Runs Qwen 3.8 27B on old V100 GPUs that do overnight work; calls it a solid worker when batched.HN comment · Sep 24, 2026

Says Qwen 3.8 delivers Opus 5-level performance with less overengineering.HN comment · Sep 22, 2026

Says Qwen 3.8 gives better results than frontier models for basic coding tasks like CRUD apps and config debugging.HN comment · Sep 22, 2026

Found Qwen 3.8 cheaper and significantly higher quality than GPT-5.6 Sol on a production task.HN comment · Sep 12, 2026

Complaints

Says Qwen 3.8 handles large-context projects poorly and drifts off track.HN comment · Nov 1, 2026

Says 4-bit Qwen 3.8 27B is iffy for coding and needs Q5 or Q6 to be useful.HN comment · Nov 1, 2026

Says Qwen 3.8 27B is barely acceptable for decompilation, overthinks, and sometimes goes off track.HN comment · Oct 24, 2026

Says quantized Qwen 3.8 models loop endlessly in thinking and are not frontier-level, though they use Qwen quants in production with thinking off.HN comment · Oct 10, 2026

Tried Qwen 3.8 on CPU/RAM; the full benchmark was estimated at 15+ hours, too slow.HN comment · Oct 5, 2026

Finds Qwen 3.8 slower and less good than Ornith 1.5 on their hardware.HN comment · Oct 4, 2026

Says Qwen 3.8 models are too large for low-spec machines, which they tested for agentic coding.HN comment · Oct 4, 2026

Gave up on Qwen 3.8 beyond planning after it ran out of context writing basic unit tests.HN comment · Sep 30, 2026

Summaries are paraphrased. Each links to the original comment.

FAQ

Is Qwen 3.8 worth it?

It depends on the job. Approval is 50%, with praise for local use and basic coding and complaints about reliability, speed, and large-context work.

Is Qwen 3.8 better than GLM 5.2?

The data is too thin to say. One developer preferred Qwen 3.8 over GLM 5.2, and one preferred GLM 5.2 instead.

Is Qwen 3.8 good for coding?

Opinions are mixed. Developers left 8 positive and 7 negative comments on code quality, and several reported poor results on large-context coding and low-bit quantized builds.