DeepSeek beats Opus in blind Blender benchmark at fraction of cost

A developer benchmarked five frontier models on rebuilding an Adidas running shoe in headless Blender from product photos. DeepSeek V4.1-Flash won at $0.23 versus Opus's $33, while transparently listing 9 unreproduced details.

2026-10-18 ~ 2026-10-18 · 3 related posts