AI Models & Platforms
Alibaba Cloud Releases Qwen4-Max with 25% Lower Singapore API Prices

Key PointsQwen4-Max cuts Singapore list prices 25% below Qwen3.8-Max for both input and output…
View more
- Qwen4-Max cuts Singapore list prices 25% below Qwen3.8-Max for both input and output tokens.
- Thinking and the final answer share one output budget so deliberation consumes response allowance.
- The subscription plan bars automation scripts and backend use leaving pay-as-you-go API calls as the alternative.
Alibaba Cloud released Qwen4-Max through Model Studio on November 16, 2026, introducing a flagship multimodal model with a one-million-token context window and lower list prices than Qwen3.8-Max.
The Singapore price list sets Qwen4-Max at $1.50 per million input tokens and $4.50 per million output tokens, compared with $2 and $6 respectively for Qwen3.8-Max. Both rates are 25% lower, calculated directly from those list prices. A one-million-token free quota is also available in Singapore for 90 days from the latest of Model Studio activation, model release or application approval.
Multimodal Input and Adjustable Reasoning
The model documentation describes text, image and video input with text output, alongside streaming, web search, parallel tool calling, function calling, prefix completion and structured JSON output. Developers can call the model through OpenAI Chat Completions, OpenAI Responses, Anthropic Messages or Alibaba Cloud’s DashScope APIs, providing several established integration routes for coding and agent software.
Alibaba Cloud designed Qwen4-Max for coding, office productivity, frontend design and complex agentic tasks. Thinking is enabled by default at high reasoning effort, with low, high and xhigh settings available. Its 131,072-token default thinking budget matches the maximum non-thinking output, while thinking and the final answer share a 262,144-token allowance, so deliberation occupies part of the same output budget as the response.
Explicit Caching and Input Costs
The Qwen4 caching guide says identical input prefixes, such as system prompts, tool descriptions and reference documents, can be reused across requests from the same user. An explicit cache’s prefix must contain at least 1,024 tokens, with five-minute or one-hour lifetimes available. Explicit breakpoints provide deterministic hits for an identical prefix within the validity period; ordinary implicit caching identifies common prefixes automatically but does not guarantee hits. Automatic caching can also place a breakpoint at the end of the input, allowing it to advance as a conversation grows.
In Singapore, cache reads cost $0.075 per million tokens for either caching approach, while explicit-cache creation costs $1.875 for a five-minute cache and $3 for a one-hour cache. Only input tokens receive the cache discount; output tokens remain at the standard price. For agents that repeatedly send long instructions or documents, prefix stability and cache lifetime therefore influence whether lower input charges accompany the reduced list rates.
Regional Access and Subscription Limits
Qwen4-Max is listed in Beijing, Singapore, Frankfurt, Virginia, Tokyo and Hong Kong. Batch inference is supported only in Beijing among those locations, while fine-tuning and custom model deployment are unsupported throughout the documented regions. API configuration, reasoning settings and caching therefore carry much of the practical control over how the managed model is used.
The full Qwen4-Max is also included in Singapore’s Personal Edition Token Plan, which bills through Credits and supports interactive tools including Claude Code, Cursor, Qwen Code, Qoder and OpenClaw. The subscription is limited to interactive coding and agent-tool use; automation scripts, custom application backends and non-interactive batch calls are prohibited. That makes the plan a distinct access route from ordinary pay-as-you-go API calls.
The full release follows Qwen4-Max-Preview, which became available at 10:00 on November 11, 2026 (UTC+8) exclusively through Token Plan. Its preview is scheduled to run through November 25, 2026, after which the preview model will be retired.
Qwen4-Max’s lower list rates and caching options address input costs for long-running agent workloads, but the final bill still depends on output length, reasoning and cache reuse. With fine-tuning and custom deployment unsupported, those usage controls and the choice between API and subscription access are central to the release’s practical impact.












