Towards AIToneBench Menu

Comparisons

Head-to-Head Model Comparisons

Last updated October 26, 2026, from board data generated at 20:15 UTC.

Compare two models on writing quality and cost, to see whether the quality gap is worth the difference in price. Written pages cover every pair among the top 25 models, the models with a configuration in the top 50 of the 141 configurations: 300 pages in total, each with every thinking level of both models, cost, the nine metric scores, a written verdict and a pick for each model. Any other pair opens in the comparison tool.

All models wrote the same 10 real scripts and were scored by the same four judges. A configuration is one model at one thinking level. The Writing Score is the 0 to 100 consensus of the judges and code, and the Writing Elo is a rating from pairwise comparisons of those scores. The 95% ranges describe individual configurations; overlap alone establishes neither a tie nor an order for a pair. The highest-ranked model outside the written match-ups is Grok 4.7, #26 of 84 models; its best configuration is #51 of 141 configurations.

Compare Any Two Models

A pair with a written match-up opens that page, and any other pair opens in the comparison tool. The configurations of a single model are compared on its thinking-levels page.

Key Match-Ups

Three written match-ups picked by rule from the current board. Picks name the configuration Louis recommends for each model.

Lab Pages: Every Configuration From One Lab

Each card names a lab's best configuration with its Writing Elo and its model's rank. The lab page ranks all of the lab's configurations and picks out the highest Writing Elo, the lowest cost within 200 Elo of the top, and the lowest cost.

Match-Ups With a Verdict (300)

No written match-up fits that filter. The comparison tool above covers every model on the board.