I didn’t realize that GPT-5.6 is basically dominating the cost/intelligence Pareto frontier right now, at least for this set of benchmarks. Otherwise it’s only Fable on the very high end and DeepSeek on the very low end. This Kimi model gets close, though.