Picking a Model From Its Decomposition

Expected prediction error on unseen data decomposes as

$\text{Expected MSE} = \text{Bias}^2 + \text{Variance} + \text{Irreducible Error}$

Two candidate models are trained on the same data-generating process, whose irreducible noise variance is $0.08$ for both.

The table reports the bias itself, not its square.

Which model has the lower expected MSE, and what is that MSE?