Best Qwen models
First-partyAlibaba's Qwen family is the broadest model lineup in open AI: frontier flagships, coding specialists, vision models, and tiny MoEs that run on a laptop. These are the best Qwen models in 2026 and what each one is actually for.
The best Qwen model overall is Qwen3.7 Max, Alibaba's flagship and the highest-scoring Chinese model on the Artificial Analysis Intelligence Index that we track (56.6). It is API only, at $0.80 per million input tokens.
For coding, the Qwen3 Coder line is the reason most people come to Qwen: Qwen3 Coder 480B is the open-weight specialist with a 1M token context window and a free tier on OpenRouter, Qwen3 Coder Next is the newer efficient line at $0.11 per million, and Qwen3 Coder 30B is the one to run locally.
For everyday volume, Qwen3.7 Plus ($0.40, 1M context) and Qwen3.5 Flash ($0.065, 1M context) cover the workhorse and budget tiers.
The top of the lineup: maximum reasoning hosted, or the 1M-context workhorse.
- 11Qwen3.7 MaxQwenThe flagship: Qwen's strongest reasoning, API only.Context262KInput$0.8/M
- 22Qwen3.7 PlusQwenThe everyday workhorse: 1M context at $0.40 per million input tokens.Context1MInput$0.32/M
The Coder line: open-weight specialists for repositories and agentic edits.
- 11Qwen3 Coder 480B A35BQwenThe open-weight coding specialist: 1M context, free tier on OpenRouter.Context1.049MInput$0.22/M
- 22Qwen3 Coder NextQwenThe newer open coding line: efficient agentic edits at $0.11 per million.Context262KInput$0.11/M
- 33Qwen3 Coder 30B A3B InstructQwenThe local pick: a fast MoE coder that fits prosumer hardware.Context160KInput$0.07/M
High-volume, chat, and vision picks from the wider family.
- 11Qwen3.5-FlashQwen1M context at $0.065 per million, for high-volume and routine steps.Context1MInput$0.065/M
- 22Qwen3 Next 80B A3B InstructQwenFast open-weight MoE with a free tier, great for chat and reasoning.Context262KInput$0.09/M
- 33Qwen3 VL 235B A22B InstructQwenThe open vision flagship: images and documents at $0.20 per million.Context262KInput$0.2/M
What is the best Qwen model in 2026?
Qwen3.7 Max is the best Qwen model overall, with the highest Intelligence Index score (56.6) of any Chinese model we track. For coding, Qwen3 Coder 480B is the open-weight specialist most people want, and Qwen3.7 Plus is the everyday workhorse with a 1M token context window at $0.40 per million input tokens.
What is the best Qwen model for coding?
Qwen3 Coder 480B A35B is the best Qwen coding model: an open-weight mixture-of-experts specialist with a 1M token context window for large repositories, and a free tier on OpenRouter to try it. Qwen3 Coder Next is the newer, cheaper line for agentic edits, and Qwen3 Coder 30B is the version that fits on local hardware.
Which Qwen models are open source?
Most of the family publishes open weights, including the entire Coder line, Qwen3 Next 80B, the Qwen3 VL vision models, and the small dense models. The exceptions are the hosted flagships: Qwen3.7 Max, Qwen3.7 Plus, and the Flash tier are API only.
Can you run Qwen models locally?
Yes. Qwen3 Coder 30B A3B is the standout local pick: a mixture-of-experts coder with only 3B active parameters that runs fast on a 24GB GPU or a high-RAM Mac. Qwen3 32B and the 30B A3B instruct models are solid local all-rounders too. The 480B coder needs server-class hardware.
Models
8Filter
First-partyUpdated
June 2026