What is Qwen Chat
Qwen is one of the most interesting AI ecosystems for teams that want both open model access and managed commercial deployment paths. Alibaba Cloud positions Qwen across chat, coding, reasoning, multilingual work, multimodal tasks, and enterprise deployment through Model Studio.
Qwen has developed from Alibaba’s foundation-model programme into a broad family that now spans chat, coding, reasoning, vision, audio, and multilingual work. That evolution matters because Qwen is no longer just a single assistant surface; it is part of a wider Alibaba Cloud platform strategy that combines open models with commercial APIs and enterprise deployment routes.
Core offerings
- Qwen chat, reasoning, coding, math, vision, audio, and omni model families.
- Official APIs through Alibaba Cloud Model Studio, including OpenAI-compatible access.
- Strong multilingual positioning with support for 119 languages and dialects.
- Hybrid “thinking” and “non-thinking” modes in newer Qwen3 models.
Pricing
Alibaba Cloud pricing depends on model and region. Current global list-price examples include qwen3.8-max and qwen3.7-max at US$1.65 per million input tokens and US$4.951 per million output tokens. The current catalogue also adds qwen3.8-omni-flash and qwen3.8-omni-flash-realtime for multimodal and realtime work; modality-specific token prices vary by region, and Qwen 3.8 Omni Flash supports automatic implicit caching. Alibaba Cloud may display different regional rates and time-limited discounts in the Model Studio console, so buyers should confirm the applicable deployment-region price.
Model footprint
Qwen’s public model spread is unusually broad. Alibaba describes flagship, reasoning, coding, math, vision, audio, and omni families, which makes Qwen more of a platform family than a single assistant.
Why select Qwen Chat
Qwen is attractive when multilingual coverage, model variety, or open-weight strategy matters. It also appeals to teams that want OpenAI-compatible APIs but more flexibility in model selection and deployment economics.
Official sources: Alibaba Cloud Model Studio overview, Qwen overview, Qwen model pricing, September 2026 audio-model price reduction.
Current models
Alibaba publishes one of the denser public pricing matrices in the market. The table below focuses on the main current international Qwen text and image lines rather than every regional or dated variant.
| Model | Context / output | Knowledge / training | Pricing / token use | Speed / notes |
|---|---|---|---|---|
Qwen 3.8 Maxqwen3.8-max | Up to 1M input tokens per request | Thinking and non-thinking modes supported Training data specifics: not publicly disclosed | Global list price: US$1.65 input / US$4.951 output per 1M tokens; regional rates vary | Newest flagship Qwen text model with context caching discounts |
Qwen 3.8 Omni Flashqwen3.8-omni-flash | Multimodal input and text/audio output | Thinking and non-thinking modes supported Training data specifics: not publicly disclosed | Pricing varies by modality and region; automatic implicit caching and Responses Session caching are supported | Current omni model for text, image, video, and audio workflows |
Qwen 3.7 Maxqwen3.7-max | Up to 1M input tokens per request | Thinking and non-thinking modes supported Training data specifics: not publicly disclosed | International list price: US$2.5 input / US$7.5 output per 1M tokens | Current flagship Qwen text model with context caching discounts |
Qwen 3 Maxqwen3-max | Tiered at 32K / 128K / 256K request bands | Thinking and non-thinking modes supported Training data specifics: not publicly disclosed | International list price: US$1.2 input / US$6 output up to 32K; US$2.4 / US$12 up to 128K; US$3 / US$15 up to 256K | Mainstream flagship tier with stronger price jumps on longer requests |
Qwen-Maxqwen-max | No tiered input bracket on the published international row | Non-thinking mode only Training data specifics: not publicly disclosed | International list price: US$1.6 input / US$6.4 output per 1M tokens | Older but still-live commercial line |
Qwen 3.1 / 3.5 vision and image linesqwen3-vl-235b-a22b-instruct, qwen-image | Vision and image generation / edit families | Training data specifics and cutoffs: not publicly disclosed | Example public prices: qwen3-vl-235b-a22b-instruct at US$0.4 input / US$1.6 output per 1M tokens in international scope; qwen-image around US$0.035 per image internationally | Qwen’s multimodal offering is now broad enough to matter as a platform, not just a chat model |