Nine Points Better And Cheaper: Where Qwen4 Max Still Falls Short
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Artificial Analysis gave Alibaba’s Qwen4 Max an Intelligence Index score of 43 on Nov. 16, compared with 34 for Qwen3.8 Max, and estimated a lower cost per index task. The result places Qwen4 Max above the comparable-model median cited by the source, but ninth among 10 models on its page; speed data and downloadable weights were unavailable at launch.

Alibaba introduced Qwen4 Max on Nov. 16, and benchmarking firm Artificial Analysis published its first results the same day: the model scored 43 on the company’s Intelligence Index, up from 34 for Qwen3.8 Max, while its estimated cost per index task was $3.27 versus $10.02. The score is an improvement over its predecessor, but Qwen4 Max ranked ninth of 10 models listed on its Artificial Analysis page; the firm had not published speed or latency data at launch.

Artificial Analysis reported an Intelligence Index score of 43 for Qwen4 Max. The source material says the index runs from 0 to 70 and puts the median for models at a similar price at 41. On the model page, the highest score among 10 listed models was 65. These figures place Qwen4 Max above the stated median but well below the top score on that page.

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

The benchmark firm estimated that completing one index task with Qwen4 Max costs $3.27, compared with $10.02 for Qwen3.8 Max (0902). Alibaba’s listed API prices are $1.50 per million input tokens and $4.50 per million output tokens. Those token rates and the index task estimate describe different cost measures and should not be treated as interchangeable.

Alibaba reported a separate improvement on its own Work360 test: Qwen4 Max scored 52.69, compared with 47.74 for Qwen3.8 Max. The company also said average cost per Work360 task fell from $4.43 to $1.67. These are vendor-reported results, while Artificial Analysis’s index and task-cost figures are the outside measurements available in the supplied launch-day material.

At a glance
reportWhen: Published November 16, 2026; launch-day…
The developmentArtificial Analysis published its first Qwen4 Max results on the model’s Nov. 16 launch day, reporting a higher index score and lower benchmark task cost than its predecessor.
Bar chart titled Ninth of ten on its own page, nine points above its predecessor. Artificial Analysis Intelligence Index for the 10 models on the Qwen4 Max page: Claude Fable 5.2 (max) 65, Gemini 4 Carbon (high) 60, Claude Opus 5.5 (max) 59, GPT-6 Astra (max) 57, GPT-6.1 Sol (max) 55, Muse Wonder Full (max) 50, Kimi K3.1 Pro (open weights) 46, Muse Wonder (max) 44, Qwen4 Max 43 (highlighted), Grok 4.7 (xhigh) 35. Below a dashed line: the predecessor Qwen3.8 Max (0902) 34
Qwen4 Max is ninth of the ten models on its Artificial Analysis page and nine points above its predecessor. Source: Artificial Analysis, model pages for Qwen4 Max and Qwen3.8 Max (0902), retrieved November 16, 2026. Chart: Thorsten Meyer AI.

Lower Costs, Uneven Standing

The results give buyers an early comparison on two practical questions: capability and cost. Qwen4 Max improved on its predecessor in the reported index and cost less per index task, which may make it worth evaluating for teams choosing among API models. The figures do not establish how it will perform on a particular company’s workload.

Its page ranking adds an important qualification. A score of 43 is above the cited comparable-model median of 41, but it ranks ninth among 10 models on that page. Buyers weighing the model against higher-scoring options should consider their task requirements and costs together, rather than reading the predecessor gain as evidence that Qwen4 Max leads the field.

Deployment choices are also limited by availability. Alibaba says Qwen4 Max is offered through its API and products, but its weights are closed. Organizations therefore cannot download and host this model based on the information provided. Alibaba says smaller Qwen4 models will be open-sourced, though it has not given a release date.

Products Worth Considering

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

As an affiliate, we earn on qualifying purchases.

From Qwen3.8 to Qwen4 Max

Alibaba introduced Qwen4 Max as the first model in its Qwen4 family. Its launch post describes a 2.4-trillion-parameter backbone, 71 billion parameters activated per token, and an additional 328 billion N-gram embedding parameters. The company describes the design as a sparse mixture of experts, in which only part of the model is activated for each token.

Alibaba also says the model accepts text, images and video, has a one-million-token context window, and is available through QwenCloud API, Token Plan and Alibaba apps including QwenWork, Qoder and QwenStudio. These are vendor descriptions; the supplied material does not include an independent verification of the architecture or context-window performance.

The launch post also lists company-run evaluations and demonstrations, including coding, secure-coding and professional-task tests. Alibaba says a secure-coding measure rose from 22.0% for Qwen3.8 Max to 31.8% for Qwen4 Max. It describes partner case studies as well. Such results and examples are vendor claims, and the supplied material does not establish that the demonstrations were controlled independent tests.

“Your new teammate has arrived.”

— Alibaba’s Qwen team

Speed, Access and Comparisons

Speed and latency were not available in Artificial Analysis’s launch-day data, so buyers cannot use the supplied results to compare response time or throughput. The source also does not provide enough information to determine how the index task-cost estimates translate to specific usage patterns.

The model’s weights are not downloadable. Alibaba says smaller Qwen4 models will be open-sourced, but no release date was provided. The supplied material also does not specify when independent evaluations of the company’s Work360 results or launch demonstrations might become available.

Finally, the cited rank reflects the 10 models shown on Qwen4 Max’s Artificial Analysis page at the time of reporting. The supplied source does not establish that this list represents every competing model or every relevant task. The score is one benchmark measure, not a complete assessment of performance.

Further Independent Testing

Further updates to Artificial Analysis’s model page may add the speed and latency figures missing on launch day. Buyers can also compare Qwen4 Max’s results on tasks that match their own needs as additional independent evaluations appear.

Alibaba has said smaller Qwen4 models will be open-sourced but has not announced a date. The company’s next concrete milestone on that front would be a release announcement with model details and access terms. Until then, Qwen4 Max remains a closed-weights offering available through Alibaba’s stated services.

Key Questions

What did Artificial Analysis report for Qwen4 Max?

It gave the model an Intelligence Index score of 43, compared with 34 for Qwen3.8 Max (0902), and estimated a cost of $3.27 per index task versus $10.02 for the predecessor.

Does Qwen4 Max rank first on the reported index?

No. The supplied material says it ranked ninth of 10 models on its Artificial Analysis page. The highest score listed there was 65.

How much does Qwen4 Max API use cost?

Alibaba lists prices of $1.50 per million input tokens and $4.50 per million output tokens. Artificial Analysis separately estimated $3.27 per index task; that is a benchmark task-cost figure, not the same measure as per-token pricing.

Can developers download Qwen4 Max’s weights?

No. The supplied material says Qwen4 Max has closed weights. Alibaba says smaller models in the Qwen4 family will be open-sourced, but it has not provided a release date.

What remains unknown about the launch results?

Artificial Analysis had not published speed or latency data at launch. Independent verification of Alibaba’s Work360 figures and demonstrations is also not included in the supplied source material.

Source: Thorsten Meyer AI

You May Also Like

AmenGate: The Moment Before the Scroll

AmenGate introduces a system-level prayer lock on iPhone that replaces distraction with meaningful prayer, aiming to reshape phone use habits rooted in faith.

Where AI Meets The Twilight Dial In An 83 Km NLC Watch

A fictional AI-built exhibition explores real noctilucent clouds with an interactive solar-depression dial. It is not a live observing service.

How Retrieval‑Augmented Generation Supercharges Large Language Models

Supercharging large language models with retrieval-augmented generation unlocks new levels of accuracy and relevance, but the true potential lies in how this technology evolves.

The license. Why the AI content market pays the brand-name corpus and strands the long tail.

Analysis of how licensing deals favor large publishers, leaving small publishers at a disadvantage in the AI content economy.