gpt-6VIDEO
GPT-6 Astra vs DeepSeek-4.1-Flash on the same prompt
On the same prompt, DeepSeek is slow and lacks the taste and intuition of a larger model, but four hours of work cost only 2 USD, and the author compares the two side by side.
Ruben Kostandyan
@ruben_kostardArchived 2026.09.15Originally posted 2026.09.15
Ruben Kostandyan @ruben_kostardOriginally posted 2026.09.15
GPT-6 Astra vs DeepSeek-4.1-Flash with same prompt. DeepSeek still takes super long time, lacks big model smell (taste, intuition at complex builds), but 4 hour work was just 2 USD!
x.com/ruben_kostard/status/2099533296237453406Archived 2026.09.15 10:10