Comparisons / Claude Opus 4.6 vs Claude Opus 4.8
AI models for coding · HN opinions Nov 2025 – Nov 2026
Claude Opus 4.6 vs Claude Opus 4.8: what developers actually say
Based on 421 opinions on Hacker News from developers who discussed these two, most of them first-hand users.
Claude Opus 4.6 comes out ahead. Approval is 63% among 287 opinions versus 49% among 134 for 4.8, and 90-day approval is 63% versus 57%. Positive shares favor 4.6 in every aspect: code quality 78/43 versus 35/27, UX 4/1 versus 1/3, cost 7/7 versus 6/10, and usage limits 4/3 versus 2/3. All five explicit head-to-head comments favor 4.6; pick 4.8 only for the newer release or 4.8-specific feedback.
When developers explicitly said they prefer one or switched: 5 picked Claude Opus 4.6, 0 picked Claude Opus 4.8.
Choose Claude Opus 4.6 if
- Want the stronger approval record, 63% versus 49%
- Prioritize code quality, autonomy, and UX
- Want the side that won all five explicit comparisons
Choose Claude Opus 4.8 if
- Want the newer release
- Need feedback specific to 4.8
- Accept a smaller evidence base for recency
Where each one wins and loses
PraiseComplaintsA = Claude Opus 4.6B = Claude Opus 4.8
Code quality
Reliability
Autonomy
Cost & pricing
Speed
Large codebases
Usage limits
UX
Why developers switched
Picked Claude Opus 4.6
Believes Opus 4.8 and Opus 5 feel worse than 4.6 after prolonged heavy use.HN comment · Aug 22, 2026
Says 4.8's safety restrictions block or weaken 5-10% of tasks, making it worse than 4.6 in practice.HN comment · Jul 23, 2026
Finds 4.8 less useful than 4.6 for multi-turn assistant work because it over-prioritizes finishing tasks.HN comment · Jul 7, 2026
Reports that 4.8 underperforms 4.6 and even Sonnet 4.6 on advanced software work.HN comment · Jun 8, 2026
Picked Claude Opus 4.8
No explicit switches recorded.
Summaries are paraphrased. Each links to the original comment.
Side by side
| Claude Opus 4.6 | Claude Opus 4.8 | |
|---|---|---|
| Approval | 63% | 49% |
| Last 90 days | 63% | 57% |
| Opinions | 287 | 134 |
| First-hand users | 242 | 117 |