Qwen2-1.5B
Alibaba · 2024-06-07 · 1.5B parameters
Qwen2-1.5B is a base (pretrained) checkpoint of Alibaba's Qwen2 series, released on June 7, 2024. It is a dense decoder-only Transformer with 1,543,714,304 parameters, grouped-query attention and tied embeddings, and Qwen documents a 32,768-token context length; Qwen2-1.5B-Instruct is its instruction-tuned counterpart.
Benchmark scores
| Benchmark | Score |
|---|---|
| C-Eval | 70.6 |
| Capability | 86.7 |
| CMMLU | 70.3 |
| GPQA - release-reported, subset unspecified | 20.7 |
| HumanEval+ | 29.9 |
| MBPP+ | 37.6 |
| MMLU STEM | 42.7 |
| MMLU-Redux | 51.8 |
| MultiPL-E (HumanEval, 8-language average) | 27.9 |
| TruthfulQA (Unspecified Metric) | 45.9 |