Benchmarks · creative
SlopBench
Runs 53 chat-style prompts and reports the percentage of model responses containing classic AI slop patterns, detected deterministically without an LLM judge. Maintained by Dan Cleary. Score: Slop Rate: percentage of responses with at least one slop-pattern hit (lower is better).
25 models · 25 results · data as of 2026-10-29 · slop-bench.vercel.app ↗
Leaderboard
GPT-5.3 Codex xhigh
All results
25 rows| # | Model | Effort | SlopBench |
|---|---|---|---|
| 1 | GPT-5.3 Codex | xhigh | 1.9% |
| 2 | Opus 4.8 | max | 3.8% |
| 3 | GPT-5.4 | xhigh | 4.3% |
| 4 | GPT-5.3 | xhigh | 5.7% |
| 5 | GPT-5.4 Mini | xhigh | 5.7% |
| 6 | GPT-5.5 | xhigh | 5.7% |
| 7 | GPT-5.4 Nano | xhigh | 7.5% |
| 8 | Sonnet 4.6 | max | 9.8% |
| 9 | Opus 4.7 | max | 11.3% |
| 10 | Grok 4.20 | xhigh | 13.2% |
| 11 | Fable 5 | max | 15.1% |
| 12 | Haiku 4.5 | xhigh | 15.1% |
| 13 | Gemini 3 Flash | high | 15.1% |
| 14 | GPT-5.4 Pro | xhigh | 17.0% |
| 15 | Grok Build 0.1 | 18.9% | |
| 16 | GLM 5 | high | 19.2% |
| 17 | Qwen 3.5 397B-A17B | max | 23.5% |
| 18 | Opus 4.6 | max | 24.5% |
| 19 | Kimi K2 | medium | 24.5% |
| 20 | Grok 4.1 Fast | high | 26.4% |
| 21 | Kimi K2.5 | max | 26.4% |
| 22 | MiniMax M2.5 | high | 26.4% |
| 23 | Gemini 3.1 Pro | high | 28.3% |
| 24 | Qwen 3 Max | max | 32.7% |
| 25 | DeepSeek V3.2 | high | 44.2% |