Qwen1.5 (0.5B to 72B; later 32B and 110B)
Qwen1.5 series from 0.5B to 72B, with chat variants and AWQ/GPTQ/GGUF quantizations, grew to 32B (Apr 1) and 110B (Apr 25).
Six base sizes (0.5B, 1.8B, 4B, 7B, 14B, 72B) with matching chat models and a broad quantization set at launch, then Qwen1.5-32B and 110B added in April. Alibaba's blog titled the 110B 'the first 100B+ model of the Qwen1.5 series'.
- Date
- Sunday, 4 February 2024
- Lab
- Alibaba (Qwen)
- Kind
- open-weights
- Access
- open weights (restricted license)
Figures
| Measure | Value | Measured by |
|---|---|---|
| Context length | 32,768 tokens for all sizes Base and chat models; Qwen1.5 code merged into Hugging Face transformers 4.37 | company |
Blog 'Introducing Qwen1.5' is dated 2024-02-04; base weights were created on Hugging Face 2024-01-22/23. Launch sizes were 0.5B, 1.8B, 4B, 7B, 14B and 72B plus the MoE; 32B (2024-04-02) and 110B (2024-04-25) followed. Benchmark tables were not opened.
Sources
- qwenlm.github.io/blog/qwen1.5/
- huggingface.co/api/models?author=Qwen&sort=createdAt&direction=1&limit=100
This record was checked against its sources on 6 October 2026. How we check