A single metric to compare open-weight and proprietary LLMs using data from public benchmarks. Every bar displays the 90% interval surrounding the score. If two bars intersect, consider those models equivalent in performance.
Loading…