Hunyuan A13B-Instruct: benchmark results
Tencent's Hunyuan-A13B, an 80B sparse MoE (13B active) instruction-tuned model (June 2025). Provider: Tencent. Released 2025-06-27. Access: Open.
Aggregate Rating 1534 ± 12, rank #666 of 1656 rated models, from 229 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AGC-Bench - futuregen | 1.52 | Dataset z-score | 98.8 |
| AGC-Bench - metaphoric_analogies | 0.63 | Dataset z-score | 81.7 |
| OpenCompass Language - Instruction Following - Chinese (CompassBench 2507) | 79.6 | Score (%) | 81.5 |
| AGC-Bench - historical_analogy | 1.05 | Dataset z-score | 81.2 |
| AGC-Bench - scimon | 0.32 | Dataset z-score | 67.9 |
| AGC-Bench - science_analogies | 0.39 | Dataset z-score | 67.5 |
| AGC-Bench - slang_generation | 0.53 | Dataset z-score | 67.1 |
| AGC-Bench - fig_qa | -0.39 | Dataset z-score | 65.9 |
| AGC-Bench - chinese_homophonic_puns | 0.52 | Dataset z-score | 65.4 |
| AGC-Bench - ocw_connections | 0.73 | Dataset z-score | 65.2 |
| ReLE - Education - Primary School Subjects | 62.7 | Accuracy (%) | 65.2 |
| Enkrypt AI - Bias Risk | 79.33 | Risk Score | 64.4 |
Interactive version: theaggregate.ai/model?slug=hunyuan-a13b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.