Open SourceAlibaba (Qwen Team)Released 2025-04

Qwen 3

Alibaba's highly capable open-weights model with top-tier multilingual performance

Context window

128K

Input / 1M tokens

Free

Output / 1M tokens

Free

Provider

Alibaba (Qwen Team)

Open weights under Apache 2.0 license. API pricing available via Alibaba Cloud DashScope.

Qwen 3 is available in sizes from 0.6B to 235B, with the flagship 235B MoE model delivering performance that rivals GPT-4o. The Qwen family excels at multilingual tasks, supporting 100+ languages with notably strong performance in Asian languages. Under Apache 2.0, the weights are truly open for commercial use and fine-tuning.

Strengths

  • Exceptional multilingual support (100+ languages)
  • Apache 2.0 license - fully open for commercial use
  • Multiple size variants from 0.6B to 235B MoE
  • Strong math and coding across models
  • Leading performance for Chinese language tasks

Best for developers who...

Multilingual applicationsSelf-hosted cost-sensitive deploymentsCustom fine-tuning on domain-specific dataAsian market applications

Benchmarks

BenchmarkScoreNotes
MMLU~87%235B MoE variant
HumanEval~89%235B MoE variant

Source: Qwen 3 official blog

Compare Qwen 3 with

All model comparisons →

Learn the concepts