Back to Qwen3
Alibaba's next-generation 80B MoE Qwen3 instruct model with 3B active params — strong reasoning and instruction-following at efficient inference cost.
131K tokensFree / Open weightsMoEApache 2.0
Benchmarks
Qwen3-Next-80B-A3B-Instruct
41.4%Qwen3-235B-A22B
52.1%Qwen3-32B
48.7%Qwen3-8B
42.6%Qwen3-30B-A3B
41.4%Qwen3-30B-A3B family FC — rank #41