Qwen3: Think Deeper, Act Faster

GitHub Pages · 1 year ago · 3

Alibaba released Qwen3, a family of large language models including a 235 billion parameter flagship model and smaller variants. Qwen3-235B-A22B achieves competitive results against DeepSeek-R1, o1, o3-mini, Grok-3, and Gemini-2.5-Pro across coding, math, and general capabilities benchmarks. The smaller Qwen3-30B-A3B model outperforms QwQ-32B while using 10 times fewer activated parameters, expanding the model family's range from 4 billion to 235 billion parameter versions.