Gemini 4 Neon leads finance eval with ~68% fewer output tokens than Fable 5.2

echen · x · 2026-11-20

Surge AI shared details from DAYJOB: Finance, an 80-task benchmark covering credit memos, valuation models, covenant tests, and stress tests. Gemini 4 Neon scored 50.5%, beating GPT 6.2 Astra (45.8%) and Claude Fable 5.2 (43.7%), while generating just 205K output tokens versus Fable's 644K — roughly one-third the tokens for a higher success rate.

Related event: Gemini 4 Neon Tops Surge Finance Benchmark at Far Lower Cost Than Claude(2 posts)→

Original post →

More from Models

Models channel →