AIToday
Large Language ModelsAI Coding AssistantsOpen-Source AIGIGAZINE AIPublished: Nov 16, 2026, 16:00 JST

Qwen4-Max: 1M-token model beats Claude Opus 5.5 on coding tests

Send on LINE
Qwen4-Max: 1M-token model beats Claude Opus 5.5 on coding tests

3 Key Points

  1. What happened

    Alibaba's Qwen team launched Qwen4-Max, a 2.4-trillion-parameter multimodal model with a 1M-token context window, scoring above Claude Opus 5.5 on several tests including DeepSWE v1.1.

  2. Why it matters

    It ranks 6th on Code Arena | WebDev, ahead of Claude Fable 5.1, and its API is priced at $1.5 per 1M input tokens — positioning it as a cheaper rival to US flagships.

  3. What to watch

    The team has said it will release open-weight versions of the Qwen4 series at a later date.

WHO IT HITSEnterprise developers and AI procurement teams choosing a model for coding agents and multimodal tasks now have a lower-cost option that trades blows with US flagship models on several benchmarks. The promised open-weight release could matter to teams that prefer to self-host models.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

Qwen4-Max arrives as Chinese labs push into the same benchmark territory as US flagships. On the Artificial Analysis Intelligence Index v5.0.1, which aggregates several benchmarks into a single score, Qwen4-Max recorded 43 — a wide gap below US flagship models like Claude Fable 5.2 and Gemini 4 Carbon, but ahead of US cost-efficiency models such as Claude Haiku 5.5 and Muse Ember. That split suggests the model is being positioned less as a raw-capability leader than as a strong performer per dollar.

The benchmark table accompanying the launch compares Qwen4-Max with Qwen3-Max-0902, Claude Opus 5.5, GPT-6.1 Sol, GPT-6 Astra, and Kimi K3.1. Beyond scores, the team points to practical tasks: operating Blender to build 3D models, and developing games tied to Unity, Unreal Engine, and Godot Engine. Those examples hint that the target is agentic and production work rather than chat alone.

The open-weight pledge is the piece with the longest reach. No date is given for that release, so how much of the model's capability reaches self-hosted deployments remains an open question.

FAQ
How much does Qwen4-Max cost to use?
API pricing is $1.5 per 1M input tokens, $4.5 per 1M output tokens, and $0.075 per 1M cached input tokens.
What can Qwen4-Max take as input?
It accepts text, images, and video, and outputs text.
Will there be an open version of Qwen4-Max?
The development team has announced it will release open-weight versions of the Qwen4 series, though no date was given.
Send on LINE

Also reported by AI Watch (Impress), SiliconANGLE AI

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleSocial Sim'27 CFP at AAAI-27 seeks work on LLM attribution problem