Gemma 4 31B beats several frontier models on the FoodTruck Bench

Gemma 4 31B takes an incredible 3rd place on FoodTruck Bench, beating GLM 5, Qwen 3.5 397B and all Claude Sonnets!

I'm looking forward to how they'll explain the result. Based on the previous models that failed to finish the run, it would seem that Gemma 4 handles long horizon tasks better and actually listens to its own advice when planning for the next day of the run.