Claude Sonnet 4.5: what developers actually say

We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on Claude Sonnet 4.5. 93 of the 97 opinions come from people who say they use it themselves.

57%approval, all opinions

–approval, last 90 days

#33of 45 models ranked

91different developers

Claude Sonnet 4.5 has 57% approval from 97 opinions by 91 authors. Code quality is split at 22 positive and 21 negative, and reliability is weaker at 5 positive and 10 negative. Praise centers on spec-driven work, scraper configs, and well-known patterns; complaints cite edit loops, subtle logic errors, and poor architecture. Approval fell from 57% in November 2025 to 33% in January 2026, then reached 80% in March, but every month in the data from April 2026 onward has only 1 or 2 opinions, so recent scores carry little weight.

Pick it if

Want help with spec-driven work, scraper configs, or one-off scripts
Work on well-known problems with established patterns
Plan to keep human review in the loop for maintenance

Look elsewhere if

Need dependable multi-step edits without loops or git restores
Expect solid project architecture and production-ready output
Cannot risk frontend performance regressions

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

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

Recalls README files generated by Sonnet 4.5 as clear and concise.HN comment · Sep 14, 2026

Says Sonnet 4.5-era models suit long-term maintenance better because humans stay in the loop.HN comment · Jul 26, 2026

Says Sonnet 4.5 was the game changer for spec-driven development coding.HN comment · Jun 13, 2026

Says Claude Sonnet 4.5 and 4.6 did most of the heavy lifting in building their Android app.HN comment · May 31, 2026

Regularly switches back to Sonnet 4.5 and prefers it over 4.6.HN comment · Apr 6, 2026

Sees a massive gap between open models and Sonnet 4.5 for C++ and Rust systems programming.HN comment · Mar 27, 2026

Says Sonnet 4.5 was very helpful but only raised their productivity about 10-15%.HN comment · Mar 11, 2026

Recommended as the model for writing scraper configs in their production setup.HN comment · Mar 3, 2026

Complaints

Says Sonnet 4.5 often loops on small edits, wasting tokens or forcing a git restore.HN comment · Feb 28, 2026

Says Sonnet 4.5 looped through increasingly convoluted regexes and ended with wrong matches.HN comment · Feb 13, 2026

Says Sonnet 4.5 called the prompt's rules nonsensical, gave normal code, and ignored the history requirement.HN comment · Jan 26, 2026

Says Sonnet 4.5 wrote a subtle logic error in their game code that only review caught.HN comment · Jan 24, 2026

Says Claude Sonnet 4.5 brought no significant improvement and is still not good enough for production.HN comment · Jan 22, 2026

Says Sonnet 4.5 suggested poor fixes for reference counting and rewrote comments during refactors.HN comment · Jan 12, 2026

Says Claude Sonnet 4.5 cut their website's mobile Lighthouse performance score from 95 to 40.HN comment · Jan 10, 2026

Says Sonnet 4.5 does a poor job with project architecture, leaving them unsure where code lives.HN comment · Jan 9, 2026

Summaries are paraphrased. Each links to the original comment.

Claude Sonnet 4.5 compared

Claude Sonnet 4.5 vs Claude Sonnet 4.6 Claude Sonnet 4.5 vs Claude Opus 4.5

FAQ

Is Claude Sonnet 4.5 worth it?

It has 57% approval from 97 opinions, so reactions are mixed. Code quality feedback is nearly even at 22 positive and 21 negative. It draws praise for spec-driven work, scraper configs, and small scripts.

Is Claude Sonnet 4.5 better than Claude Sonnet 4.6?

No clear winner appears in the data. Two developers preferred Sonnet 4.5 over 4.6, and two preferred 4.6 instead.

Is Claude Sonnet 4.5 reliable for production code?

Reliability feedback is negative overall at 5 positive and 10 negative mentions. Complaints include edit loops, subtle logic errors, weak architecture, and a mobile Lighthouse score drop from 95 to 40. One developer called it not good enough for production, and others describe errors that review caught.