Write docs & copy
Pick the best model for docs and copy from human-like writing and creative-writing Elo.
Right now GPT-6 Astra is first on 1 of 2 boards it enters.
Sorted by compositeSynced 2026-10-10
| GPT-6 Astra | 85 | 2/2 | 29 | 1 |
| GPT 6 Sol | 84 | 2/2 | 20 | 5 |
| IInkling Small | 83 | 1/2 | 1 | — |
| DeepSeek V4 Flash | 79 | 1/2 | 5 | — |
| Claude Haiku 5.5 | 79 | 1/2 | 4 | — |
| MMercury 2.5 | 79 | 1/2 | 2 | — |
| GPT-6 Sol Pro | 78 | 1/2 | 7 | — |
| GPT-6.1 Sol Pro | 76 | 1/2 | 10 | — |
| Grok 4.6 | 76 | 1/2 | 9 | — |
| DeepSeek V4 Pro | 76 | 1/2 | 8 | — |
| Claude Fable 5.1 | 75 | 2/2 | 49 | 3 |
| Claude Sonnet 5.5 | 75 | 1/2 | 12 | — |
| GPT 5.6 Sol | 74 | 2/2 | 11 | 12 |
| Claude Opus 5 | 73 | 2/2 | 46 | 4 |
| Claude Opus 5.5 | 73 | 2/2 | 27 | 8 |
| GLM-5.3 | 72 | 2/2 | 35 | 7 |
| NNemotron 3.5 Lightning | 72 | 1/2 | 15 | — |
| GPT-6 Luna Pro | 72 | 1/2 | 14 | — |
| MiniMax M3 | 71 | 1/2 | 17 | — |
| GLM-5.3 Prime | 71 | 1/2 | 16 | — |
Why it matters
Human-like ≠ strong creative prose
Slop Index asks how human email/tweet copy feels; Creative Writing ranks short creative prose by Elo. The leaders are not the same set.
Least slop is not the priciest
After inversion, GPT-5.6 Luna leads Slop with a source Slop Score of 26. Pricier Fable 5 ($30/M) sits higher on the source board—more slop, not less.
People and machines both vote
Slop Index: 40% live blind arena plus 60% mechanical axes. Creative Writing: rubric absolute scores plus pairwise Elo, judged by Claude models.