Tim Street

Carbohydrate estimates from food photographs by Claude Fable 5.1 and GPT-6 Astra: a September 2026 update to a reproducibility benchmark

In April 2026 a benchmark sent 13 food photographs about 500 times each to four vision-capable large language models (LLMs) at temperature 0.01 and reported median within-photograph coefficients of variation (CV) of 2.4% to 11.0%. Two newer models, Claude Fable 5.1 and GPT-6 Astra, accept no sampling parameter and cann...