WHAT NEW IN X11.03 TUE

AI collections from X · 22 collections · 65,263 posts

gpt-6POST

Blogger calls GPT-6 Astra the first AGI model and questions benchmark accuracy

The blogger says they test models day and night, finds DeepSeek 4.1 Flash far behind GPT-6 Astra, and argues the benchmark chart does not reflect the real gap.

Derya Unutmaz, MD
@DeryaTR_Archived 2026.09.16Originally posted 2026.09.16
Derya Unutmaz, MD @DeryaTR_Originally posted 2026.09.16
Open original on X

I'm a big supporter of open-source models, but DeepSeek 4.1 Flash is nowhere near the quality of GPT-6 Astra, which is the first AGI model, so not even close! It seems benchmarks aren't reflecting this. I try these AI models day and night, and this graphic is simply not true.

This post quotes another post: “DeepSeek V4.1 Flash said to cost 1% of GPT-6 Astra to train”

x.com/deryatr_/status/2099900753494098314Archived 2026.09.16 10:21