🔍 Read the full analysis: How Qwen4 Max Balances A Nine-Point Gain With A Lower Price on ThorstenMeyerAI.com
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
TL;DR
Artificial Analysis gave Alibaba’s Qwen4 Max a score of 43 on its Intelligence Index, nine points above Qwen3.8 Max (0902), and calculated a $3.27 cost per index task versus $10.02 for the predecessor. The new model ranks ninth among 10 models on its comparison page; it is available through Alibaba services, but its weights are closed and launch-day speed data was unavailable.
Alibaba’s new Qwen4 Max scored 43 on Artificial Analysis’s Intelligence Index, nine points above its predecessor, while the benchmark firm estimated that an index task costs $3.27 to run, compared with $10.02 for Qwen3.8 Max (0902). The independent results, published on November 16, show a measured improvement and lower task cost, but Qwen4 Max ranked ninth among the 10 models on its comparison page and had no published speed or latency figures.
Artificial Analysis tested Qwen4 Max at its highest reasoning setting, xhigh, and reported an Intelligence Index score of 43. The index combines nine evaluations covering areas including agentic coding, office automation, knowledge and physics. The firm described 43 as above the 41-point median among comparable models. Alibaba’s API default is high rather than xhigh, so the result does not directly establish how the model performs under its default setting.
The same benchmark page places Qwen4 Max ninth of 10 models. Its listed leading scores include 65 for Claude Fable 5.2 (max), 60 for Gemini 4 Carbon (high) and 59 for Claude Opus 5.5 (max). Qwen4 Max’s score is therefore a clear increase over Qwen3.8 Max’s 34, but it does not put the new model near the top of that displayed comparison.
Artificial Analysis calculated a cost of $3.27 per index task for Qwen4 Max, against $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. The benchmark’s task-cost figure is distinct from those per-token prices: it reflects the estimated usage required for its index tasks, not a flat charge for every request.
Qwen4 Max · launch day
From launch claim to independent check
Alibaba’s numbers come first. An outside measurement shows where the model actually lands.
- The launchAlibaba calls Qwen4 Max a 24/7 “teammate”, with 2.4 trillion parameters and a one-million-token context window.
- The vendor’s own testOn its in-house Work360 benchmark, Alibaba reports 52.69, up from 47.74 for Qwen3.8 Max.
- The independent indexArtificial Analysis scores it 43, up from 34 and just above the median of 41 for models at a similar price.
- 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 at Lower Task Cost
The results give developers and businesses a first outside measure of the trade-off Alibaba is pitching: higher benchmark performance at a lower measured task cost than its previous flagship. That could make Qwen4 Max worth evaluating for workloads where the model’s quality is adequate and inference spending matters. The numbers are not proof that it is the best-value option for every use: costs depend on the task, token use, reasoning setting and provider pricing.
Its position on the comparison page also tempers the improvement. A nine-point gain is meaningful against Qwen3.8 Max, but the overall 43-point score remains below the page’s leading models. Buyers need to weigh that gap against the price, their own accuracy requirements and the model’s performance on their particular work rather than treating a single index as a universal ranking.
Access is another consideration. Qwen4 Max is available through Alibaba’s services but its weights are not downloadable, limiting independent deployment and external inspection. That distinction matters for organizations that require self-hosting, model customization or direct review of model parameters.
Products Worth Considering

LG UltraGear 32GX870A-B 32-inch 4K UHD OLED Dual-Mode Gaming Monitor

Bose QuietComfort Headphones, White Smoke

Samsung 32-Inch Odyssey G5 QHD Gaming Monitor (G50F)
As an affiliate, we earn on qualifying purchases.
What Alibaba Says It Built
Alibaba introduced Qwen4 Max as the first model in its new Qwen4 family. Its launch blog describes a 2.4-trillion-parameter backbone, 71 billion parameters activated per token and an additional 328 billion N-gram Embedding parameters. The design is a sparse mixture-of-experts architecture: according to the company, only part of the network is active for each token. Alibaba also lists changes to attention and residual-layer design intended to support long inputs and training stability.
Those architecture and training details are Alibaba’s disclosures, not independently verified findings, because the company has not released the weights. Alibaba says the model handles text, images and video and has reasoning enabled by default at high effort. It is offered through QwenCloud API, the Token Plan subscription and Alibaba applications including QwenWork, Qoder and QwenStudio. The Next Web reported that a limited preview began on November 11.
Alibaba also reported an improvement on its in-house Work360 benchmark: a score of 52.69, up from 47.74 for Qwen3.8 Max. The company said average Work360 cost per task fell from $4.43 to $1.67. These are vendor-reported figures from Alibaba’s own evaluation, separate from Artificial Analysis’s independent index and cost estimate. The launch materials also describe coding results and customer examples, but those claims should not be treated as independently established outcomes.
“In our Work360 evaluation, the average cost per task drops from $4.43 to $1.67 — a 62% reduction.”
— Alibaba
Speed, Defaults and Independent Checks
Artificial Analysis had no speed or latency data for Qwen4 Max on launch day. The available results therefore do not show how quickly it responds or how throughput compares with other models—factors that can affect both user experience and operating costs.
The main benchmark score was measured at xhigh reasoning effort, while Alibaba’s API default is high. The source material does not provide an independent score at that default setting, so the headline result may not represent a typical API configuration. It also does not establish how the model performs across individual tasks or under different prompts.
Alibaba’s Work360 and coding claims remain vendor-reported, while customer examples are case studies rather than controlled comparisons. The company has not released model weights, and it has not given a date for open-sourcing smaller Qwen4 models. The benchmark and launch material available here do not settle how Qwen4 Max performs in long-term production use or what costs customers will see across varied workloads.
More Testing Will Settle the Trade-Off
The next useful evidence will be independent testing at Alibaba’s default reasoning setting, alongside measurements of latency, throughput and cost across real-world tasks. Comparisons using the same prompts and settings could clarify whether the lower index-task cost holds for particular business workloads and how much performance changes from xhigh to high.
Customers can access Qwen4 Max through Alibaba’s API and listed products now, according to the launch information, but decisions about deployment will depend on their own evaluations. Alibaba has said smaller Qwen4 models will be open-sourced, but has not announced a release date. Until then, users seeking downloadable weights will need to consider other models.
Key Questions
What did Artificial Analysis score Qwen4 Max?
Artificial Analysis gave Qwen4 Max 43 points on its Intelligence Index after testing it at xhigh reasoning effort. The firm said the score was above the 41-point median among comparable models.
How does Qwen4 Max compare with its predecessor?
Qwen4 Max scored 43, compared with 34 for Qwen3.8 Max (0902), a nine-point difference on the same index. Artificial Analysis also estimated a lower cost per index task: $3.27 versus $10.02.
Is Qwen4 Max the top-ranked model in the comparison?
No. It ranks ninth among 10 models on the Artificial Analysis page for Qwen4 Max. Its score is above the comparable-model median, but below the listed leading models.
Can users download Qwen4 Max?
No. Qwen4 Max is offered through Alibaba’s services and its weights are closed. Alibaba says smaller models in the Qwen4 family will be open-sourced, but has not provided a date.
Are Qwen4 Max’s speed and Alibaba’s benchmark claims independently confirmed?
Artificial Analysis had no speed or latency data at launch. The Intelligence Index result is an external measurement, but figures from Alibaba’s Work360 benchmark and its customer examples are company-reported and are not independent confirmation.
Source: ThorstenMeyerAI.com
