Overview
Qwen2.5 was released in September 2024 with seven sizes from 0.5B to 72B, plus specialized coder and math variants. It became one of the most popular open bases for fine-tuning.
Highlights
- Seven sizes from 0.5B to 72B
- Pretrained on about 18 trillion tokens
- Strong structured output (JSON) and long-text generation
- Most sizes under Apache 2.0
Typical uses
Fine-tuning, local assistants, data extraction and multilingual applications.
Things to consider
- The 3B and 72B models use the Qwen license rather than Apache 2.0.
- Superseded by Qwen3 for most new projects.
- Download sizes refer to the default 4-bit quantized versions in the Ollama library; full-precision weights are larger.
Specifications are taken from the model card on Hugging Face, the Ollama library and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.