benchmark
.
space
benchmarks
rankings
compare
voices
transcripts
papers
articles
YouTube · 2026-04-08
"Deepseek V3 model already shocked the world last year by demonstrating GPT4 class performance at roughly 1/27th the inference cost of OpenAI offerings. V4 takes that to another level."
Alex
Co-host, Asia Tech Macro
DeepSeek V3.2
view original source →
all researcher takes →