A Third Of The Cost, But Not Yet At The Top: Qwen4 Max’s Results
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: A Third Of The Cost, But Not Yet At The Top: Qwen4 Max’s Results on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Alibaba launched Qwen4 Max on November 16, 2026, describing a 2.4-trillion-parameter model with vision and a one-million-token context window. Artificial Analysis gave it an Intelligence Index score of 43, nine points above its predecessor, while estimating an index-task cost of $3.27; it ranked ninth among 10 models on its page. The model is available through Alibaba services, but its weights and launch-day speed data were unavailable.

Alibaba launched Qwen4 Max on November 16, 2026, and independent benchmark firm Artificial Analysis published its first results the same day: the model scored 43 on the Intelligence Index, up from 34 for its predecessor, while an index task was estimated to cost $3.27. The score is above the 41-point median among comparable models, but Qwen4 Max ranked ninth among 10 models listed on its Artificial Analysis page, underscoring that a lower bill does not put it at the top of that comparison.

Artificial Analysis tested Qwen4 Max at its highest reasoning setting, xhigh, using Intelligence Index v5.0.1, which combines nine evaluations covering areas such as agentic coding, office automation, knowledge and physics. Alibaba’s API default is high reasoning effort, so the published score reflects a more intensive setting than the default. Artificial Analysis describes 43 as above average among comparable models, whose median is 41.

The benchmark firm estimated the cost of running one Intelligence Index task at $3.27, compared with $10.02 for Qwen3.8 Max (0902). Alibaba lists API prices of $1.50 per million input tokens and $4.50 per million output tokens. These are separate measures: the index-task estimate is not a universal price for every user’s workload, which depends on token use and task setup.

Alibaba says Qwen4 Max has a 2.4-trillion-parameter backbone, with 71 billion parameters activated per token, plus 328 billion N-gram Embedding parameters. The company describes it as a sparse mixture-of-experts model and says it can process text, images and video, with a one-million-token context window. The model is offered through QwenCloud API and Alibaba products including QwenWork, Qoder and QwenStudio; its weights are not available to download.

At a glance
reportWhen: Launched and independently benchmarked…
The developmentArtificial Analysis published its first benchmark results for Alibaba’s Qwen4 Max on launch day, showing a higher score and lower per-task cost than its predecessor, alongside a ninth-place ranking among 10 listed models.

Qwen4 Max · launch day

From launch claim to independent check

Alibaba’s numbers come first. An outside measurement shows where the model actually lands.

Alibaba: 2.4T parameters, 71B active Own test Work360: 47.74 → 52.69 Independent index (0 to 70) 0 70 34 41 43 65 Speed: no data yet
  1. The launchAlibaba calls Qwen4 Max a 24/7 “teammate”, with 2.4 trillion parameters and a one-million-token context window.
  2. The vendor’s own testOn its in-house Work360 benchmark, Alibaba reports 52.69, up from 47.74 for Qwen3.8 Max.
  3. The independent indexArtificial Analysis scores it 43, up from 34 and just above the median of 41 for models at a similar price.
  4. What is still openThe top model on the same page scores 65. Speed is not measured yet, and the weights are closed.

Sources: Alibaba/Qwen launch blog, Nov 16, 2026 (vendor figures); Artificial Analysis, model pages for Qwen4 Max and Qwen3.8 Max (0902), retrieved Nov 16, 2026. Schematic scale; marker positions follow the index values.

Graphic: Powered by Thorsten Meyer AI

Lower Costs, Still Behind Leaders

The results put a measurable figure on two different parts of the launch pitch: Qwen4 Max improves on the prior flagship, and Artificial Analysis estimates that its index run costs substantially less. For companies comparing hosted models, per-task expense can matter alongside benchmark performance, especially when a system is used repeatedly. The reported cost is tied to a specific evaluation, however, and should not be treated as a price guarantee across different workloads.

Its ranking supplies an equally important qualification. On the same Artificial Analysis page, the top score is 65; several other listed models score between 55 and 60. A 43 therefore indicates a gain over Qwen’s previous model, not leadership on this index. The figures help separate a substantial generational improvement from a claim that the model has overtaken the strongest listed competitors.

Access also affects how customers can assess the product. Qwen4 Max is currently API-only and closed-weight, so outside researchers cannot inspect or independently run the model weights. Buyers must rely on hosted access and published tests unless Alibaba later releases weights or additional technical material.

Products Worth Considering

Samsung Galaxy Tab A11+ 11-Inch Tablet, 6GB RAM, 128GB Storage, Expandable, Gray

Samsung Galaxy Tab A11+ 11-Inch Tablet, 6GB RAM, 128GB Storage, Expandable, Gray

Bose QuietComfort Wireless Noise-Cancelling Headphones, Petal Pink

Bose QuietComfort Wireless Noise-Cancelling Headphones, Petal Pink

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.

What Alibaba’s Launch Adds

Alibaba introduced Qwen4 Max as the first model in its Qwen4 family, calling it a “24/7 AI teammate” for work, creativity and discovery. The launch announcement outlined changes to attention and model architecture, including Gated DeltaNet layers, Qwen Sparse Attention and parallel residual branches. Alibaba also said the model completed training without loss spikes. These are company descriptions; because the weights are closed, the architecture and training claims cannot be independently checked from the model itself.

Alibaba’s own Work360 evaluation reported a score increase from 47.74 to 52.69 compared with Qwen3.8 Max. The company said average cost per task on that benchmark fell from $4.43 to $1.67, a 62% reduction. Work360 is Alibaba’s in-house test, so those results are vendor-reported rather than independent confirmation. The company also described coding, manufacturing and other demonstrations as case studies, not controlled comparisons.

Other reported benchmark figures offer a mixed picture. The Next Web, citing FourWeekMBA, reported a score of 75.7 on DeepSWE v1.1, slightly above Claude Opus 5.5, but a Terminal-Bench 4.0 score of 48.5, below Claude Opus 5.5 and Kimi K3.1. Those numbers were attributed to Alibaba’s chart rather than an independent test. Artificial Analysis’s published index is the clearest outside comparison available in the supplied launch-day material.

“In our Work360 evaluation, the average cost per task drops from $4.43 to $1.67 — a 62% reduction.”

— Alibaba

Open Questions on Access and Speed

Artificial Analysis had no speed or latency data for Qwen4 Max on launch day, so users cannot yet compare how quickly it responds with the figures provided. The independent index result was run at xhigh, while Alibaba’s API default is high; the source material does not provide a corresponding independent score at the default setting.

It is also unclear when, or whether, Qwen4 Max weights will become available. Alibaba said that smaller Qwen4 models will be open-sourced, but did not provide a release date, and that statement does not promise an open-weight version of Max. The available reporting also does not establish how the model performs across a broad set of real customer workloads, or whether Alibaba’s in-house savings figures will hold outside Work360.

Awaiting More Independent Tests

The next useful evidence will be independent results at multiple reasoning settings, alongside published latency and throughput measurements. Those would help customers judge whether the launch-day index score reflects everyday API use and how the model’s cost compares for specific tasks, rather than for one benchmark run.

Readers should also watch for any release schedule for the smaller Qwen4 models Alibaba says it plans to open-source, and for further documentation about Qwen4 Max’s capabilities and evaluation methods. Until then, the confirmed picture is a model that scores higher and costs less per index task than its predecessor, but remains below the leading scores on the cited comparison and unavailable as downloadable weights.

Key Questions

How did Qwen4 Max score on Artificial Analysis?

It scored 43 on Intelligence Index v5.0.1, tested at the xhigh reasoning setting. Artificial Analysis gave the predecessor, Qwen3.8 Max (0902), a score of 34.

Is Qwen4 Max the highest-ranked model in the comparison?

No. It ranked ninth among 10 models on its Artificial Analysis page. The highest score shown there was 65.

How much does an index task cost?

Artificial Analysis estimated $3.27 per Intelligence Index task, versus $10.02 for Qwen3.8 Max. Alibaba’s listed API rates are $1.50 per million input tokens and $4.50 per million output tokens; actual usage costs vary by workload.

Can users download Qwen4 Max?

No. Qwen4 Max is closed-weight and offered through Alibaba’s API and products. Alibaba said smaller Qwen4 models would be open-sourced but gave no date.

Are speed and latency results available?

Not in Artificial Analysis’s launch-day results. The firm reported no speed or latency data for Qwen4 Max at that time.

Source: ThorstenMeyerAI.com

You May Also Like

Podman V6.0.0

The latest version of Podman, v6.0.0, has been officially released, introducing new features and performance enhancements for container management.

OpenAI Models Are Coming To Google Cloud: What To Know

OpenAI says its models are coming to Google Cloud, but has not disclosed the lineup, launch date, pricing or access details.

OpenAI’s Cursor Disabling: Developer Collateral In AI Innovation

OpenAI plans to disable models for Cursor, an AI coding tool now owned by SpaceX, citing trust and contractual concerns amid corporate shifts.

Signal: Four Frontier-Class Open Models in Eight Weeks — China’s Release Cadence Is the Story

Chinese AI labs released four frontier-class open models within eight weeks, marking a rapid production line that challenges Western efforts.