Key Specifications
| Vendor | alibaba |
|---|
| Version | 2.5-1-5b |
|---|
| Release Date | 2024-09-19 |
|---|
| Context Window | 32768 tokens |
|---|
| Input Modalities | text |
|---|
| Output Modalities | text |
|---|
| License | Qwen License |
|---|
| Documentation | https://qwenlm.github.io/ |
|---|
Benchmark Performance
| Benchmark | Score | Unit | Evaluated At | Notes | Source |
|---|
| MMLU | 54.5 | % | 2024-09-19 | 5-shot | view |
| HUMANEVAL | 35.2 | pass@1 | 2024-09-19 | — | view |
| GSM8K | 33.2 | % | 2024-09-19 | 0-shot CoT | view |
| MATH | 9.5 | % | 2024-09-19 | 0-shot CoT | view |
| BBH | 51.5 | % | 2024-09-19 | 3-shot CoT | view |
| GPQA | 20.2 | % | 2024-09-19 | 0-shot | view |
| IFEVAL | 41.9 | % | 2024-09-19 | prompt_strict | view |
| ARC | 76.7 | % | 2024-09-19 | challenge | view |
| MUSR | 37.8 | % | 2024-09-19 | 0-shot | view |
| WINOGRANDE | 64.4 | % | 2024-09-19 | 0-shot | view |
Compliance
- Data Residency: CN
- SOC2: ✗
- HIPAA: ✗
- GDPR: ✗
- ISO 27001: ✗
Qwen2.5 1.5B
Ikhtisar Model
阿里巴巴 Qwen2.5 1.5B 超轻量开源模型, 32K 上下文, 1.5B 参数, 适合资源受限设备部署。
Spesifikasi Inti
| Vendor | Versi | Tanggal Rilis | Jendela Konteks | Modalitas Input | Modalitas Output | Lisensi |
|---|
| Alibaba | 2.5-1-5b | 2024-09-19 | 32K | text | text | Qwen License |
Kinerja Benchmark
| Benchmark | Skor | Satuan | Catatan |
|---|
| MMLU (Massive Multitask Language Understanding) | 54.5 | % | 5-shot |
| HumanEval | 35.2 | pass@1 | — |
| GSM8K (Grade School Math 8K) | 33.2 | % | 0-shot CoT |
| MATH | 9.5 | % | 0-shot CoT |
| BBH (BIG-Bench Hard) | 51.5 | % | 3-shot CoT |
| GPQA | 20.2 | % | 0-shot |
| IFEval | 41.9 | % | prompt_strict |
| ARC | 76.7 | % | challenge |
| MUSR | 37.8 | % | 0-shot |
| WinoGrande | 64.4 | % | 0-shot |
Harga
| Input | Output | Baca Cache | Tulis Cache |
|---|
| — | — | — | — |
per juta token
Kelebihan
Kekurangan
- MMLU 仅 54.5，知识推理偏弱。
- HumanEval 35.2，代码能力较弱。
- 闭源专有模型，不支持自托管。
Kasus Penggunaan
Referensi