Key Notes
- Alibaba has released Qwen4-Max for coding and long-running agents, expanding beyond the earlier limited preview.
- Standard API rates are $1.50 per million input tokens and $4.50 per million output tokens, with a one-million-token context window.
- Selected coding benchmarks show gains, but independent evaluations do not place Qwen4-Max ahead of every Claude or Kimi model.
Alibaba has launched Qwen4-Max, its new flagship AI model for coding, professional work and long-running agents, intensifying competition with Anthropic and Moonshot AI. The November 16 release brings broader developer access and lower token prices, alongside benchmark results that show gains on selected tasks rather than a clean sweep of its rivals.
The Qwen team announced a model with native visual understanding, a one-million-token context window and a new mixture-of-experts architecture. Its backbone contains 2.4 trillion parameters, with 71 billion activated per token. Alibaba positions it as a system for carrying projects through multiple steps, including work in design software and game engines.
Strong Coding Results, With Clear Gaps Elsewhere
Alibaba’s benchmark graphic shows Qwen4-Max scoring 75.7 on DeepSWE v1.1, an agentic coding evaluation, narrowly ahead of Claude Opus 5.5 at 74.2 and Kimi K3.1 at 75.5 in that comparison.
The same chart illustrates why the result needs context. On Terminal-Bench 4.0, Qwen4-Max scores 48.5, compared with 59.6 for Opus 5.5 and 67.2 for Kimi. These are different tasks and testing setups; a lead on one evaluation does not establish that a model is stronger across software development.
Independent results also give a more measured view. As of November 16, Artificial Analysis lists Qwen4-Max at 43 on its current Intelligence Index, compared with 46 for Kimi K3.1 Pro and 65 for Claude Fable 5.2 at their highest listed settings. The index combines several evaluations and measures something broader than a single coding benchmark.
The comparison also moves quickly: Anthropic released Fable 5.2 last week. Comparisons with Fable 5.1 therefore describe an older version of Anthropic’s model.
API Access Starts at $1.50 In and $4.50 Out
QwenCloud’s Singapore-region model page lists standard prices of $1.50 per million input tokens and $4.50 per million output tokens. Implicit cache-hit input costs $0.075 per million tokens. Explicit cache creation has separate charges.
The model accepts text, images and video, produces text, and supports web search, tool calls and structured outputs. Developers can use compatible OpenAI and Anthropic API formats.
Alibaba opened a preview on November 11, when developer access was restricted to Token Plan subscriptions. The launch adds ordinary API access; Token Plan remains another route, alongside QwenStudio, QwenWork and Qoder.
An Agent That Delivers Work in Stages
The Qwen team has also published its Live Delivery plugin. It encourages agents to produce a usable first version, then deliver improvements while users review progress and change requirements. The initial release supports Codex CLI on Linux and macOS; it does not include a model or API key.
Downloadable models remain a separate question. Alibaba promises open weights for smaller members of the Qwen4 family, without naming a release date. Qwen4-Max itself is being offered through hosted services.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.