LiveCodeBench hard

Stored LiveCodeBench hard accuracy from the tool-calling feed. 28 models have a score. The highest is 71.14 for O4-Mini (High). The lowest is 1.89 for Claude-3-Haiku. Bar length is that number on a 0 to 100 scale. A higher number is a longer bar. This is not the overall pass at 1 and not the easy or medium split. A missing number stays off the chart.