Qwen
Qwen3 8B
Qwen3 8B is a text model with thinking and non-thinking modes. Native context is 32,768; the 40,960 configuration value and the YaRN extension to 131,072 are separate properties.
Specifications
- Source-reported parameters
- 8,190,735,360
- Architectures
- Unknown
- Layers
- 36
- KV heads
- 8
- Head dimension
- 128
- config.json context
- 40960
- Native context in model card
- 32768
- Extended context
- 131072
- Weight dtype
- Unknown
- File formats
- Unknown
- Declared languages
- Unknown
- Estimation method
- Weights + logical KV, excluding runtime