Solved to an exploitability of 1 milli-big-blind / game, which is stated to be within ~approximately optimal over the bounds of a human lifetime of play.