AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Qwen4 Max Benchmark Update: Nine Points Better And Still Catching Up 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

Artificial Analysis’s first results for Alibaba’s Qwen4 Max put it at 43 on the Intelligence Index, nine points above Qwen3.8 Max (0902), and above the comparable-model median of 41. It ranks ninth among 10 models on its page; independent speed data was not available, and Alibaba’s benchmark and case-study figures remain company-reported.

Artificial Analysis published its first benchmark results for Alibaba’s Qwen4 Max on Monday, November 16, giving it a score of 43 on the Intelligence Index. That is nine points above Qwen3.8 Max (0902), but the new model ranks ninth among 10 models listed on its Artificial Analysis page, placing the result between a measurable improvement and a substantial gap to the leaders.

Artificial Analysis says the score is above the median of 41 for models in a similar price range. The index combines nine evaluations covering areas including agentic coding, office tasks, knowledge and physics. The company tested Qwen4 Max at its highest reasoning setting, xhigh; Alibaba’s API default is reported as high, so results from default use may differ. These are initial results, published on the model’s launch day.

The same Artificial Analysis model page lists Qwen4 Max below nine other models, with a top score of 65. Examples on that page include Claude Fable 5.2 (max) at 65, Gemini 4 Carbon (high) at 60 and Claude Opus 5.5 (max) at 59. Kimi K3.1 Pro, identified in the source as the strongest Chinese model on the chart, scores 46. These comparisons describe the models and settings shown on that page; they do not establish how every model performs in every task.

Artificial Analysis estimates the cost of running one index task at $3.27 for Qwen4 Max, compared with $10.02 for Qwen3.8 Max. Alibaba’s listed API prices are $1.50 per million input tokens and $4.50 per million output tokens. The vendor separately reports a drop in average Work360 task cost from $4.43 to $1.67, but that figure comes from Alibaba’s own evaluation, not the independent index.

At a glance
updateWhen: Published November 16, 2026; launch-day…
The developmentArtificial Analysis published its first Qwen4 Max results on November 16, reporting a nine-point gain over its predecessor and a ninth-place ranking among 10 models on the model page.

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

A Better Score, Still Below Leaders

The results give buyers an early independent point of comparison for Alibaba’s new flagship. A nine-point increase over its predecessor suggests a meaningful improvement on this particular index, while the score of 43 and ninth-place position show that the gains do not put it near the top of the listed field. The distinction matters for organizations weighing a model’s advertised capabilities against external measurements.

Cost could also shape that decision. Artificial Analysis’s estimate of $3.27 per index task is lower than the predecessor’s $10.02 under its test, though task costs depend on usage and evaluation setup. The lower estimate does not by itself show that Qwen4 Max is cheaper for every workload, nor that its outputs are equivalent across tasks. Buyers will need to compare quality, latency, reliability and pricing on their own use cases.

The evidence is incomplete on two fronts. The benchmark is a single early snapshot, and speed and latency data were unavailable from Artificial Analysis at launch. The independent score therefore helps place the model, but does not settle how it performs in routine production use.

Products Worth Considering

LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1, Black

LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1, Black

Kindle Paperwhite Signature Edition 32GB (2024 Model) - Metallic Black

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

As an affiliate, we earn on qualifying purchases.

Alibaba’s Launch and Test Claims

Alibaba introduced Qwen4 Max as the first model in its Qwen4 family, describing it as a “24/7 AI teammate” for work, creativity and discovery. The company says it has a 2.4-trillion-parameter backbone, 71 billion activated parameters per token and an additional 328 billion N-gram embedding parameters. It also describes the system as supporting text, images and video, with a one-million-token context window. These are launch details and specifications supplied by Alibaba; the model’s weights are not available for outside inspection.

Alibaba’s launch blog reports a Work360 score of 52.69, up from 47.74 for Qwen3.8 Max, and says average cost per task on that test fell from $4.43 to $1.67. It also reports that the share of functions passing both functionality and security checks on a secure-coding benchmark rose from 22.0% to 31.8%. Those figures are vendor-reported. Alibaba’s described manufacturing, legal and research examples are case studies, not controlled independent tests.

Some additional benchmark figures have been reported by The Next Web, citing FourWeekMBA: Qwen4 Max scored 75.7 on DeepSWE v1.1, compared with 74.2 for Claude Opus 5.5, while scoring 48.5 on Terminal-Bench 4.0, below Opus 5.5 at 59.6 and Kimi K3.1 at 67.2. These are reported figures from the cited coverage, and the source material notes that Alibaba’s comparison chart was published as an image and was not independently transcribed here.

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

— Alibaba

Speed, Weights and Real-World Results

Independent latency and speed results were not available on launch day, so users cannot yet compare response times from the supplied measurements. The model’s weights are also closed: Qwen4 Max is available through Alibaba’s services rather than as a downloadable model. Alibaba says smaller Qwen4 models will be open-sourced, but the source material gives no release date.

The index results are early and setting-dependent. Artificial Analysis used xhigh reasoning effort, while the API default is high; the supplied information does not quantify how much that setting changes the score. It is also not yet clear how the model performs across a wider range of independent tests or in sustained customer use. Alibaba’s Work360, coding and case-study results have not been independently verified in the source material.

More Independent Testing Needed

The next useful developments are additional independent benchmark results, especially measurements of speed, latency and performance at the API’s default reasoning setting. Comparisons across more tasks and real workloads would help determine whether the nine-point improvement translates into consistent gains for users.

Readers can also watch for updates to Artificial Analysis’s model page and for Alibaba to provide a date or details for the planned open release of smaller Qwen4 models. Until then, Qwen4 Max remains a closed, API-accessed model with an encouraging predecessor comparison but a ranking that leaves it well behind the leading entries on the cited index.

Key Questions

What score did Qwen4 Max receive?

Artificial Analysis scored it 43 on its Intelligence Index, tested at xhigh reasoning effort. The company says that is above the comparable-model median of 41.

How much did it improve over the previous model?

Qwen4 Max scored nine points higher than Qwen3.8 Max (0902), which scored 34 on the same index, according to the model pages cited in the source material.

Does the score make Qwen4 Max a top-ranked model?

No. It ranks ninth among 10 models on its Artificial Analysis page, whose highest listed score is 65. Its result is above the stated comparable-model median, but below the leading entries shown.

Is Qwen4 Max open source?

No downloadable weights are available for Qwen4 Max. Alibaba says smaller models in the Qwen4 family will be open-sourced, but no date is provided in the source material.

What performance information is still missing?

Artificial Analysis had no speed or latency data for the model on launch day. The available results also do not establish performance across a broad range of independent real-world tests.

Source: ThorstenMeyerAI.com

You May Also Like

Creating An Image-Free Floating-Seam Journey: The KIRIN EXPRESS Approach

An AI-built showcase uses code-drawn scenery, scroll-driven motion and interactive controls to present a fictional 603 km/h maglev journey.

A Timeline Check On Musk’s Promises For Tesla Robotaxis

Elon Musk said Tesla will wind down Cybertruck production at its Texas plant around mid-2027 to make room for more Cybercab production.

What To Know Before Using Selfies To Track Skincare Results

A proposed skincare app would track skin changes in weekly selfies, but its metrics and product recommendations still need validation.

NVIDIA Runs Simulation Of Fault-Tolerant Quantum Computing At Jülich Supercomputing Centre

NVIDIA says its 8,192-GPU JUPITER simulation modeled fault-tolerant quantum computing with realistic hardware noise and tested magic-state cultivation.