SlopBench
Slop so hard mfs tryna find me.
Slop Rate
Percentage of outputs containing classic AI slop. Lower is better.
| # | Model | Slop ↑ |
|---|---|---|
| 1 |  openai/gpt-5.3-codex | 1.9% |
| 2 |  anthropic/claude-opus-4.8 | 3.8% |
| 3 | openai/gpt-5.4 | 4.3% |
| 4 | openai/gpt-5.3-chat | 5.7% |
| 5 | openai/gpt-5.4-mini | 5.7% |
| 6 | openai/gpt-5.5 | 5.7% |
| 7 | openai/gpt-5.4-nano | 7.5% |
| 8 | anthropic/claude-sonnet-4.6 | 9.8% |
| 9 | anthropic/claude-opus-4.7 | 11.3% |
| 10 |  x-ai/grok-4.20-beta | 13.2% |
| 11 |  google/gemini-3-flash-preview | 15.1% |
| 12 | anthropic/claude-haiku-4-5 | 15.1% |
| 13 | anthropic/claude-fable-5 | 15.1% |
| 14 | openai/gpt-5.4-pro | 17.0% |
| 15 | x-ai/grok-build-0.1 | 18.9% |
| 16 |  z-ai/glm-5 | 19.2% |
| 17 |  qwen/qwen3.5-397b-a17b | 23.5% |
| 18 |  moonshotai/kimi-k2-thinking | 24.5% |
| 19 | anthropic/claude-opus-4.6 | 24.5% |
| 20 | x-ai/grok-4.1-fast | 26.4% |
| 21 |  minimax/minimax-m2.5 | 26.4% |
| 22 | moonshotai/kimi-k2.5 | 26.4% |
| 23 | google/gemini-3.1-pro-preview | 28.3% |
| 24 | qwen/qwen3-max | 32.7% |
| 25 |  deepseek/deepseek-v3.2 | 44.2% |