Qwen3

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Qwen3 — the entire lineup from 0.6B to 235B-A22B under Apache 2.0, about 36T tokens, 119 languages

Why it matters

A strategy of covering every size class with an unrestricted license overtook Llama in the number of derivative models on HF

Source — Primary sources first; secondary reports are labelled “Press”

Officialqwenlm.github.iohttps://qwenlm.github.io/blog/qwen3/

Added in Week 35, 2026 · See that week

On the map — ⑨ Open-Weight Lineage line

  1. Previous stopKanana Nano 2.1B
  2. This recordQwen3
  3. Next stopQwen3-Coder
See it on the map

In the model family tree — Qwen family · 13 versions

  1. Earlier versionQwQ-32B
  2. This recordQwen3
  3. Later versionQwen3-Coder
See the whole Qwen family

Same week

Cohere Labs and others publish The Leaderboard IllusionPaperMem0 publishes its paperPaper

Nearby on the same axis

Naver open-sources HyperCLOVA X SEEDOfficialTII releases Falcon ArabicOfficialMeta releases Llama 4 Scout and Maverick amid LMArena submission controversyPressKakao open-sources Kanana 1.5 under Apache 2.0Official