DevConsensus
Models / Claude Sonnet 5
Model · HN opinions Nov 2025 – Nov 2026

Claude Sonnet 5: what developers actually say

We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on Claude Sonnet 5. 33 of the 38 opinions come from people who say they use it themselves.

61%approval, all opinions
41%approval, last 90 days
#35of 45 models ranked
37different developers

Claude Sonnet 5 has 61% approval across 38 opinions from 37 authors, a mildly positive but split result. The 90-day score is 41%, and the monthly trend fell from 75% in June and 76% in July to 36% in August, then rose to 50% in September. Code quality is the main strength at 9 positive and 5 negative mentions, and autonomy leans positive at 5 to 2. Cost is 3 positive and 5 negative, and speed is 0 positive and 3 negative; the monthly samples are small.

Pick it if

  • Work on large repos and need strong code quality
  • Need an agent that persists when a task stalls
  • Need reliable instruction following in long contexts

Look elsewhere if

  • Need low latency or fast turnaround
  • Need predictable token costs
  • Do most of your coding inside an IDE on a budget

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
5
9
Cost & pricing
5
3
Autonomy
2
5
Reliability
2
4
Speed
3

How opinion has moved

050100Jun '26Jul '26Aug '26Sep '26
Monthly approval (line) and number of opinions (bars). Months with fewer than 3 opinions are left out.

What developers say

Praise

Scored higher than Sonnet 5.5 on their esoteric-language benchmark, persisting more instead of asking questions.HN comment · Sep 28, 2026
Says Claude Sonnet 5 finished the same cleanup task that DeepSeek V4 Pro stalled on.HN comment · Sep 9, 2026
Using Sonnet 5 on a large porting project and says it is doing a good job.HN comment · Aug 13, 2026
Says Chinese models felt significantly dumber than Claude Sonnet 5 when tried in Pi.HN comment · Aug 11, 2026
Calls Sonnet 5 at low effort a workhorse that is resourceful when things go wrong and does not cheat.HN comment · Jul 29, 2026
Switched from Opus to Sonnet 5 and sees no quality drop on large repos.HN comment · Jul 24, 2026
Says Sonnet 5 correctly found an unhandled retry-and-failure bug that GPT-5.6 Terra missed.HN comment · Jul 18, 2026
Says fresh Sonnet 5 holds conventions and instructions much better than Opus 4.8 in long contexts.HN comment · Jul 18, 2026

Complaints

Says Sonnet 5 burned $400 on one incident and needed many prompts to tune alerts.HN comment · Sep 29, 2026
Says Sonnet 5 is not usable for coding.HN comment · Sep 22, 2026
Says Sonnet 5 went on a long side quest through Git history for a one-line doc change.HN comment · Aug 24, 2026
Says Sonnet 5 trusted a pasted error and chased a wrong lead, ignoring the user's mistake.HN comment · Aug 18, 2026
Says Sonnet 5 was great at first but now feels slow and weaker.HN comment · Aug 14, 2026
Says Sonnet 5 is slightly better on long tasks but nearly unusable in-IDE, slow and costly with no caching.HN comment · Aug 8, 2026
Says Sonnet 5 is nearly unusable for coding after getting used to Opus 4.8.HN comment · Jul 13, 2026
Says Anthropic's current tokenizer produces many more tokens than OpenAI's on the same code, inflating costs.HN comment · Jul 13, 2026

Summaries are paraphrased. Each links to the original comment.

FAQ

Is Claude Sonnet 5 worth it?

The data is mixed. Approval is 61% overall and 41% over the last 90 days. Code quality and autonomy draw praise, while cost and speed draw criticism.

Is Claude Sonnet 5 better than Claude Opus 4.8?

Opinions are split. One developer preferred Sonnet 5 over Opus 4.8, and two preferred Opus 4.8. Some say Sonnet 5 is hard to use for coding after Opus 4.8, but one reports no quality drop after switching on large repos.

Is Claude Sonnet 5 good for coding?

Code quality is the most discussed aspect, with 9 positive and 5 negative opinions. Two developers say it does not work well for coding, and speed has 0 positive and 3 negative mentions. The full sample is 38 opinions.