Qwen reportedly estimates food macros nearly 8x faster than GPT Astra in one run
A user asked three AI models to estimate macronutrients, such as protein, from the same food photo. They say all three returned estimates around 1,200–1,300 calories.
TLDR
Frustrated with time-consuming food logging, a longtime macro tracker says they tested three models on the same food photo. They reported response times of 2.46 seconds for Qwen 3.8 27B on Cerebras, 19.55 seconds for OpenAI’s GPT Astra and 14.74 seconds for Anthropic’s Fable 5.1. According to the user, all three estimated around 1,200–1,300 calories, and Qwen finished nearly 8x faster than Astra in that run.
Combined views
28.9K
2 Sources, first seen ago