🔍 Read the full analysis: Qwen4 Max Cuts The Cost To A Third, Yet The Leaders Keep Their Lead on ThorstenMeyerAI.com
Get smart everyday buys 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
Alibaba launched Qwen4 Max on Nov. 16, 2026, and Artificial Analysis’s first results put it at 43 on the Intelligence Index, nine points above its predecessor. The benchmark’s estimated cost per task was $3.27, compared with $10.02 for the previous model, but Qwen4 Max ranked ninth among 10 models on its comparison page. Alibaba’s separate performance and cost figures are company-reported, and independent speed data was not available at launch.
Alibaba launched Qwen4 Max on Nov. 16, 2026, as the first model in its Qwen4 family, while independent benchmark results showed a 43-point Intelligence Index score and an estimated $3.27 cost per task. Artificial Analysis said the score was above the median for comparable models and nine points higher than the previous Qwen flagship, but Qwen4 Max ranked ninth among 10 models listed on its comparison page.
Artificial Analysis tested Qwen4 Max at its highest reasoning setting, xhigh. Its Intelligence Index version 5.0.1 combines nine evaluations spanning areas such as agentic coding, office automation, knowledge and physics. The benchmark’s model page lists Qwen4 Max at 43 points, compared with 34 for Qwen3.8 Max (0902). Artificial Analysis described 43 as above the median of 41 among models in a similar price range.
The same comparison shows a substantial gap between Qwen4 Max and the highest-scoring systems on that page. Claude Fable 5.2 (max) scored 65, Gemini 4 Carbon (high) 60, Claude Opus 5.5 (max) 59, GPT-6 Astra (max) 57 and GPT-6.1 Sol (max) 55. Kimi K3.1 Pro, identified in the source as the strongest Chinese model on the chart, scored 46. Those results describe the models and settings listed on the Artificial Analysis page; they do not establish performance across every task or deployment.
Artificial Analysis estimated a cost of $3.27 for one index task with Qwen4 Max, versus $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 benchmark-cost figure and the token prices are different measures: the former estimates the cost of running the benchmark task, while the latter are the company’s listed rates.
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
Lower Benchmark Costs, Lower Ranking
The launch gives buyers an early comparison between performance and cost, rather than relying only on Alibaba’s launch claims. The independent results indicate a measurable improvement over Qwen3.8 Max at a lower estimated benchmark cost. For teams comparing API models, that may make Qwen4 Max worth testing where price matters and the task fits its capabilities.
But the score also puts a limit on the launch-day takeaway. Qwen4 Max was ninth of 10 models on its Artificial Analysis page, with a nine-point gap to the listed leader. A lower cost per benchmark task does not by itself show that the model is more cost-effective for every workload: real costs can depend on prompt length, output length, reasoning settings and how often a task must be repeated.
Access also shapes who can evaluate the model. Qwen4 Max is available through Alibaba’s API and products, but its weights are not downloadable. That means outside developers cannot inspect or run the full model independently on their own infrastructure, limiting direct verification of its architecture and performance.
Popular right now

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

Bose QuietComfort Wireless Noise-Cancelling Headphones, Sandstone

Dyson Airwrap Co-anda2x Multi-Styler and Dryer (2026 Model)
As an affiliate, we earn on qualifying purchases.
A New Model in Alibaba’s Qwen Family
Alibaba describes Qwen4 Max as a model with a 2.4-trillion-parameter backbone, 71 billion parameters activated per token and an additional 328 billion N-gram Embedding parameters. The company says it uses a sparse mixture-of-experts design, in which only part of the network is activated for each token. Alibaba also describes changes to attention and training architecture. Those technical details come from the company; because the weights have not been released, outside researchers cannot independently inspect them.
Alibaba says the model accepts text, images and video and produces text. Its documentation sets reasoning to high by default, while Artificial Analysis used xhigh for its test. This difference matters when comparing scores: the published benchmark result reflects a higher reasoning setting than the stated API default, and may not match what a user sees without changing settings.
Alibaba also reported results from its in-house Work360 evaluation, which it says covers 26 Tier-1 domains, 182 Tier-2 subdomains and 721 Tier-3 fields. The company reported a score increase from 47.74 to 52.69 and an average task-cost decrease from $4.43 to $1.67. These are vendor-reported results, not independent measurements. Alibaba also described coding and business case studies, including a manufacturing programming task; the material provided does not establish that those demonstrations were controlled, independently audited tests.
“In our Work360 evaluation, the average cost per task drops from $4.43 to $1.67 — a 62% reduction.”
— Alibaba, describing its Work360 results
Questions Around Cost and Testing
Launch-day evidence remains limited. Artificial Analysis had no speed or latency data for Qwen4 Max, so users do not yet have an independent basis for comparing response time with rival models. The available material also does not provide enough information to determine how results vary across common workloads or reasoning settings.
Alibaba’s Work360, coding and case-study claims have not been independently verified in the supplied material. Its launch comparison chart was published as an image, and the source report says it did not reconstruct values from that image. Secondary reports cited by the source give mixed coding results, but those figures are attributed to other publications and do not replace a full, independently reviewed benchmark set.
It is also unclear when Alibaba will release smaller Qwen4 models with open weights. The company said those models would be open-sourced but gave no release date. Qwen4 Max itself remains closed-weight and accessible through Alibaba services.
More Independent Results Needed
The next useful evidence will be additional independent testing, particularly on latency, real-world task costs and performance at the API’s default high reasoning setting. Results across a wider range of workloads could show whether the benchmark-cost advantage persists when users run their own applications.
For now, Qwen4 Max is available through Alibaba’s QwenCloud API, Token Plan subscription and the company’s QwenWork, Qoder and QwenStudio products, according to the source material. Alibaba has said smaller models in the Qwen4 family are planned for open release, but has not announced a date. Until further tests or releases appear, the launch-day picture is a stronger score than Qwen3.8 Max, lower measured benchmark cost and a ranking below the leading models on the cited comparison page.
Key Questions
What did Artificial Analysis score Qwen4 Max?
Artificial Analysis gave Qwen4 Max 43 points on its Intelligence Index version 5.0.1, testing it at the xhigh reasoning setting. The provider listed a median score of 41 for comparable models.
How does Qwen4 Max compare with its predecessor?
Qwen4 Max scored 43, compared with 34 for Qwen3.8 Max (0902), and its estimated cost for one Artificial Analysis index task was $3.27 versus $10.02. The benchmark figures do not show how every individual user workload will compare.
Is Qwen4 Max the leading model in the cited comparison?
No. It ranked ninth among 10 models on the Artificial Analysis Qwen4 Max comparison page. The highest score listed there was 65.
Can developers download Qwen4 Max?
No. The source says Qwen4 Max is closed-weight and available through Alibaba’s API and products. Alibaba said smaller Qwen4 models would be open-sourced, but did not give a release date.
Are speed and latency results available?
Not in the cited launch-day Artificial Analysis results. The source reports that independent speed and latency data was not available for Qwen4 Max at launch.
Source: ThorstenMeyerAI.com
