In the past decade Pgt race Ug99 and its variants have been a challenge to wheat production in Kenya. Towards identifying suitable varieties, 37 lines selected from rust screening nurseries and 3 checks were tested for yield and adult plant reaction to natural stem rust epidemics across 11 diverse Kenyan environments in 2013 and 2014. Trial locations were chosen to mainly represent key wheat growing areas as well as three new sites. Evaluations based on the AMMI linear-bilinear model indicated significant (P≤0.01) genotype (G), environment (E), and GE interactions with the first three principal components (PC) explaining ~70% of the observed variation. With a contribution of over 90% to total sum of squares, environment was the predominant source of variation and the genotypic effect was approximately twofold higher than the GE effect. Based on biplot projections, clusters of lines were most closely associated with specific environments. Biplots also pointed to at least five environments, clearly those in traditional wheat growing areas that were highly correlated and associated with positive PC suggesting a similar ability to discriminate genotypes. Each non-traditional testing environment was associated with negative PC and was uncorrelated in its discriminatory ability. Combined yield and stability results achieved through classifying genotypes based on Shukla’s stability variance and Kang’s stability rating, revealed four genotypes (R1357, R1362, R1372, and R1374) as desirable candidates. The hitherto popular variety Robin, used as the ‘best check’ for yield, posted an at least 10% lower yield relative to the highest yielding genotype (R1357). Moreover, Robin which was released as a high yielding variety with adult plant resistance in 2009, was not stable in performance across environments, perhaps due its current susceptibility to a new Pgt race (TTKTT) within the race Ug99 group, that is virulent to the SrTmp-based resistance.