Updated Nov 2026 · 20,064 comments read
Which AI coding tool do developers actually rate?
Benchmarks measure models in a lab. Reviews are often paid. We read what working developers say about AI coding tools in Hacker News comments, Nov 2025 – Nov 2026, and turned 13,197 opinions into rankings.
AI coding tools
| # | Tool | Type | Approval | Last 90 days | Opinions |
|---|---|---|---|---|---|
| 1 | Pi Open source | Terminal agent | 84% | 80% | 423 |
| 2 | OpenAI Codex OpenAI | Terminal agent | 69% | 63% | 2,013 |
| 3 | OpenCode Open source | Terminal agent | 66% | 55% | 780 |
| 4 | Zed Zed Industries | Editor | 59% | 57% | 457 |
| 5 | Claude Code Anthropic | Terminal agent | 54% | 51% | 3,026 |
| 6 | Cursor Anysphere | AI IDE | 55% | 53% | 766 |
| 7 | Antigravity Google | AI IDE | 45% | 42% | 301 |
| 8 | GitHub Copilot GitHub | IDE extension | 42% | 38% | 659 |
| 9 | Gemini CLI Google | Terminal agent | 29% | 25% | 211 |
Tools with at least 100 opinions.
Most-compared
Models for coding
| # | Model | Approval | Last 90 days | Opinions |
|---|---|---|---|---|
| 1 | DeepSeek V4 Flash | 83% | 80% | 151 |
| 2 | Claude Opus 4.5 | 78% | 90% | 322 |
| 3 | Claude Opus 5.5 | 75% | 75% | 186 |
| 4 | GLM 5.2 | 75% | 82% | 142 |
| 5 | DeepSeek V4 Pro | 75% | 67% | 121 |
| 6 | GPT-5.4 | 72% | – | 114 |
| 7 | GPT-5.5 | 69% | 67% | 236 |
| 8 | GPT-5.6 Sol | 67% | 66% | 247 |
| 9 | Claude Opus 4.6 | 63% | 63% | 287 |
| 10 | Claude Fable 5 | 57% | 48% | 188 |
| 11 | Claude Opus 4.8 | 49% | 57% | 134 |
| 12 | Claude Opus 4.7 | 42% | – | 145 |
How the ranking works
Every comment is tagged with the tool it talks about, whether the opinion is positive or negative, whether the author uses the tool, and which topics it covers. Tools are ranked by approval, adjusted for sample size. Full methodology →