Qwen3

Alibaba's ultra-compact 0.6B Qwen3 base model — remarkably capable for its size, ideal for on-device and embedded applications.

Alibaba's 235B MoE Qwen3 flagship with 22B active params — top open-source performance on reasoning, coding, and math.

Alibaba Qwen's 235B instruct MoE model — high-capacity instruct model for multi-turn dialogue and large-context reasoning.

Alibaba's 30B MoE Qwen3 base model with 3B active params — strong reasoning at low inference cost for the Qwen3 family.

Alibaba Qwen's 30B instruct model — large-context, instruction-tuned model for complex reasoning and long-form generation.

Alibaba's 32B Qwen3 dense base model — strong reasoning and multilingual capabilities with hybrid thinking mode support at a practical scale.

Alibaba Qwen's 4B instruct model — compact, efficient instruction-tuned model for cost-sensitive chat and generation.

Alibaba's next-generation 80B MoE Qwen3 instruct model with 3B active params — strong reasoning and instruction-following at efficient inference cost.

Alibaba's 8B Qwen3 base model — strong reasoning and multilingual capabilities with hybrid thinking mode support.