Benchmarks
Benchmarks
SuperCLUE
Knowledgeunit % · normalized over [40, 90]Comprehensive Chinese-language LLM evaluation (CLUEbenchmark), covering reasoning, knowledge, and generation.
#ModelSourceScoreNormalized
Score distribution
6 tracked results across the normalization window
4090
Score vs. parameters
Open-weights models, log-x params
Only one open-weights model with a disclosed parameter count — not enough to plot a trend.