Kimi K3 vs Claude Opus 4.8, one shot each
We ran Moonshot's new Kimi K3 and Claude Opus 4.8 through the same 34 one-shot prompts, same harness, and they finish within a hair of each other — K3 at 3.24 average, Opus at 3.12. The interesting part is where each one misses.
K3's edge is the hard graphics work. It was the only model in the set to nail fluid-simulation, and it cleared reaction-diffusion and the Wolfenstein raycaster where Opus returned a broken or blank screen. Opus is the steadier hand on game logic and the DOM: it kept 2048 and Game of Life fully playable while K3 stalled on both, and it recovered a spinning cube that K3 left as a black screen.
The season tally and the head-to-head prompts are below — each build links to its live artifact and the grade behind it.
