Dr Pia Model Bake-off

Cosmetic Dentistry NYC, content test v2, Oct 18, 2026

Round B: Optimizer

Every optimizer starts from the same draft (Sonnet, run 1) and tries to lift the score without breaking the facts.

Leaderboard

Grok 4.78 articles96.5spread 3.959.169.8pending37.21.80.208/88/82$0.00
Astra18 articles91.0spread 2.159.174.8pending75.73.60.0018/1818/180$0.00
Solhigh18 articles90.7spread 1.758.971.9pending80.34.00.0018/1818/180$0.00
Claude Opus 5.518 articles89.4spread 3.863.075.3pending76.63.50.0018/1818/180$0.00
GPT 6.1 Sol18 articles88.2spread 3.058.971.2pending78.73.80.0018/1818/180$0.00
MiniMax M316 articles, 1 empty85.6spread 3.160.474.4pending71.23.00.0016/1610/161$0.00
DeepSeek V4 Pro16 articles, 6 empty85.2spread 2.960.277.6pending71.33.10.2016/161/1625$0.35
Terra18 articles84.5spread 4.261.773.8pending73.03.30.0018/1813/180$0.00
Claude Sonnet 5.518 articles84.4spread 3.159.374.2pending79.23.90.0018/1818/180$0.00

Click a column heading to sort. Pending means the score has not landed yet. It is never counted as zero. NeuronWriter, Surfer and Frase are shown side by side on purpose.

NeuronWriter, Surfer SEO, Frase and blind quality per model
Scores side by side

No bars yet for Frase: those scores are pending, not zero.