XuanYuan-70B — benchmark results

Du Xiaoman's Chinese financial LLM, a continued pretrain of Llama 2 70B on Chinese and finance corpora with context extended to 8k/16k (September 2023). Provider: Du Xiaoman. Released 2023-09-22. Access: Open.

Unified ELO 1458 ± 15, rank #963 of 1776 rated models, from 9 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open Chinese LLM - WinoGrande68.27Accuracy (%)86.5
Open Chinese LLM - HellaSwag68.09Accuracy (%)86.4
Open Chinese LLM - ARC Challenge59.39Accuracy (%)84.1
Open Chinese LLM - C-Eval Semantic86.68Accuracy (%)81
Open Chinese LLM Leaderboard64.23Average Score (%)78.6
Open Chinese LLM - CMMLU63.99Accuracy (%)75.2
Open Chinese LLM - GSM8K51.63Accuracy (%)70.3
CMMLU71.15-shot Avg Accuracy (%)55
Open Chinese LLM - TruthfulQA MC51.53Accuracy (%)29.1

Interactive version: theaggregate.ai/model?slug=xuanyuan-70b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.