Loading page content
Qwen3-4BOpen
Tool use · Reasoning
Evaluation overview
Overall rank
#136
58.4 points · 180 models
Open-weight rank
#68
64 points · 109 models
Data coverage
122 items
2 sources · updated through 2025-07-01
Domain performance
Coding#118 / 160
63.3
Research#129 / 174
69.1
Agents#103 / 125
56.4
Long context#82 / 82
39.4
Evaluation highlights
All evaluationsCoding · Qwen3-4B-Instruct-2507 Technical Report · Initial April 2025 model · non-reasoning mode
66.6%
7/ 31
Instruction following · Qwen3-4B-Instruct-2507 Technical Report · Initial April 2025 model · non-reasoning mode
81.2%
25/ 30
Basic information
- Context (input / output)
- 131K / 33K
- Parameters
- 4B
- Release date
- 2025-04
- License
- Apache 2.0
- Input modalities
- Text
- Output modalities
- Text
Availability
vLLM · FP8Deployment options
FP8 · 5 GB VRAM
Related links
Version relationships
Qwen3-4B family
Qwen3-4BCurrent
2025-04 · Current model
Qwen3-4B-Base
2025-04 · Same series
Qwen3-4B-Instruct-2507
2025-07 · Same series
Qwen3-4B-Thinking-2507
2025-07 · Same series