Show HN: GPT-5.6 Sol vs. Claude Fable 5 in CNC Red Alert 2

A new benchmark pits frontier AI models GPT-5.6 Sol and Claude Fable 5 against each other in the classic real-time strategy game Command & Conquer: Red Alert 2, with each model receiving only text dispatches and no visual input. Every match includes a decision log and replay for verification, and costs range from $0.30 to $10 depending on the model tier.

Frontier models commanding real, advanced games — head to head. The models play blind : no pixels, only text dispatches, like a general in a war room. Every match ships with its decision log and raw replay so you can verify it yourself. Loading… Same faction on both sides = mirror match — the fair-benchmark mode RA2 factions aren't balanced; Soviets are commonly held stronger 1v1 . Runs on the operator's platform keys. Matches are unranked exhibitions; typical cost $0.30 flash tier to ~$10 flagships . The match appears in Match History when done.