Hunyuan-1.8B-Instruct: benchmark results
Tencent Hunyuan dense instruction checkpoint with grouped-query attention and switchable fast and slow thinking. The publisher supports a 256K context; pretraining checkpoints and quantized releases remain separate. Provider: Tencent. Released 2025-07-30. Access: Open.
Aggregate Rating 1255 ± 32, rank #9207 of 10916 rated models, from 143 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CSSBench - Adult Content | 52.67 | Attack success rate (%; share of obfuscated malicious adult | 100 |
| CSSBench - Fraud and Hate | 34 | Attack success rate (%; share of obfuscated malicious fraud | 100 |
| CSSBench - Illegal Activities and Compliance | 27.33 | Attack success rate (%; share of obfuscated malicious illega | 100 |
| CSSBench - Public and Political Safety | 32.41 | Attack success rate (%; share of obfuscated malicious public | 94.4 |
| AI Energy Score (Reasoning) | 5 | Energy Score (1-5) | 82.4 |
| AI Energy Score (Text Generation) | 5 | Energy Score (1-5) | 79.5 |
| EuroEval Spanish NLU - ScaLA ES | 5.87 | Matthews correlation coefficient (%) | 40.1 |
| EuroEval Spanish NLU - Sentiment Headlines ES | 33.93 | Matthews correlation coefficient (%) | 39.6 |
| EuroEval English NLU - CoNLL EN | 59.33 | Micro F1 excluding MISC (%) | 34 |
| EuroEval Faroese NLU - FoSent | 7.84 | Matthews correlation coefficient (%) | 29.3 |
| EuroEval English NLU - ScaLA EN | 26.46 | Matthews correlation coefficient (%) | 29.1 |
| EuroEval Italian NLU - Sentipolc16 | 37.28 | Matthews correlation coefficient (%) | 27.5 |
Interactive version: theaggregate.ai/model?slug=hunyuan-1-8b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.