Qwen 3
Alibaba's open-weight Qwen 3 family covers everything from a tiny 0.6B model to a 235B mixture-of-experts. Every size has a switch that turns "thinking" mode on or off — same weights, two behaviors. Speaks 119 languages and was the first big open release to match DeepSeek-R1 on reasoning benchmarks. 阿里开放权重的 Qwen 3 家族，规格从 0.6B 一路开到 235B 的 MoE 全覆盖。每个规格都内置"思考模式"开关——同一份 权重，两套行为。支持 119 种语言，也是第一个在推理基准 上匹敌 DeepSeek-R1 的大型开源版本。
How are Intelligence, Speed & Cost bucketed?
- Top 1%≤ 1%
- Top 5%≤ 5%
- Top 10%≤ 10%
- Good≤ 25%
- Medium≤ 50%
- Below avg> 50%
- Top 1%≥ 345 tok/s
- Top 5%≥ 237 tok/s
- Top 10%≥ 196 tok/s
- Good≥ 146 tok/s
- Medium≥ 90 tok/s
- Slow< 90 tok/s
- Freeopen weights · self-host
- Low< $1 / M out
- Moderate$1–5 / M out
- High≥ $5 / M out
Why it matters
Demonstrated that the open-weight Chinese ecosystem could match Western closed reasoning models within months of o3 and R1, on permissive licenses, across the entire size spectrum from edge (0.6B) to frontier (235B MoE).
证明了开源中文生态在 o3 和 R1 发布几个月内就能跟上 西方闭源推理模型——还顶着宽松许可，从 0.6B 边缘端到 235B MoE 前沿端全尺寸段都不缺席。
Core Capabilities
Context Window
Availability
Pricing Model
Capability / Performance
Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).
What it feels like
- Best open-source reasoning model at its release — 235B-A22B (Thinking) beats DeepSeek-R1 on 17/23 benchmarks
- Toggle-able thinking mode: same weights serve both reasoning and fast-chat modes
- Strong 119-language coverage; the most genuinely multilingual frontier-tier model
- Coder variant reaches 77.2% on SWE-bench Verified — competitive with Claude 4.5 Opus's 80.9%
- GPQA Diamond 87.8% and AIME26 94.1% — frontier reasoning at open-weights pricing
- Apache-2.0 license + 1M-context coder variant make it production-ready, not a research toy
Best use cases
- Multilingual production deployments (119 languages) where most models stay English-centric
- Self-hosted reasoning workflows that need both 'fast mode' and 'thinking mode' from one weight set
- Open-weights agentic coding (Qwen3-Coder) with very large context windows
- Cost-sensitive bulk reasoning that would be prohibitive on closed APIs
Tools to try
Not ideal for
- Multimodal tasks (Qwen3 base is text — vision lives in Qwen3-VL, audio in Qwen3-Audio)
- Edge / single-consumer-GPU deployments at the 235B scale
- Workflows where Western-platform compatibility is a contractual requirement
Model Evolution
qwen is Alibaba's language model family.