🔍 Read the full analysis: Is Qwen4 Max Worth It? Nine Points Better At A Third Of The Price 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 November 16, 2026, with a 2.4-trillion-parameter backbone and a one-million-token context window. Artificial Analysis gave it a score of 43, nine points above its predecessor, and reported a $3.27 cost for an index task versus $10.02 for the earlier model. It ranked ninth of 10 models on its comparison page, and speed data was unavailable at launch.
Alibaba launched Qwen4 Max on November 16, and independent benchmark company Artificial Analysis published its first results the same day: the model scored 43 on its Intelligence Index, nine points above its predecessor, while an index task cost $3.27 compared with $10.02 for the earlier model. The results show a measurable improvement and lower benchmark cost, but Qwen4 Max ranked ninth among 10 models on the company’s comparison page; speed data was not available at launch.
Alibaba describes Qwen4 Max as the first model in its new Qwen4 family and pitches it as a “24/7 AI teammate.” The company says it has a 2.4-trillion-parameter backbone, with 71 billion parameters activated per token, plus 328 billion N-gram embedding parameters. It supports text, images and video, and has a one-million-token context window. Those specifications and capabilities come from Alibaba; the model’s weights are not public, so outside researchers cannot inspect or independently verify its architecture.
Artificial Analysis tested Qwen4 Max at its highest reasoning setting, called xhigh. Its Intelligence Index v5.0.1 combines nine evaluations, including agentic coding, office automation, knowledge and physics. The service scored Qwen3.8 Max (0902), the previous flagship, at 34. The new model’s 43-point score was above the 41 median for comparable models cited by Artificial Analysis, but below several higher-ranked systems. The top model on the Qwen4 Max comparison page scored 65.
The benchmark provider also reported a $3.27 cost per index task for Qwen4 Max, compared with $10.02 for Qwen3.8 Max. Alibaba’s published prices are $1.50 per million input tokens and $4.50 per million output tokens. These task-cost figures are tied to the benchmark’s evaluation; they are not a universal price for every user request, which can vary with input, output and usage settings. Alibaba’s own Work360 figures also show a cost reduction, but those are company-reported results rather than an independent test.
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
For businesses comparing models, the results put cost and capability in tension. Qwen4 Max improved on its predecessor and cost less per Artificial Analysis index task, which may make it worth testing for workloads where its performance is sufficient. But its ninth-place position on the comparison page signals that a lower bill does not, on these results alone, make it a leader across the index.
The score also depends on how the model is run. Artificial Analysis used xhigh reasoning effort, while Alibaba’s API default is high. A company evaluating Qwen4 Max in its own application may therefore see different performance, cost and response times depending on settings and task mix. The reported benchmark cost should be treated as one comparison point, not a promise about a customer’s bill or results.
Access and control are another part of the decision. Qwen4 Max is available through Alibaba’s API and services, but its weights cannot be downloaded. That limits deployment choices for organisations that need to run a model on their own infrastructure or inspect its internals. Alibaba says smaller Qwen4 models will be open-sourced, but has not announced a release date.
Products Worth Considering

Logitech MX Master 4 Ergonomic Wireless Mouse with Haptics – Graphite

Dell XPS Tower 8960 Desktop, Intel Core i7-13700, NVIDIA GeForce RTX 4060, 16GB DDR5, 512GB SSD

Samsung 27-Inch Odyssey OLED G5 QHD Gaming Monitor (G50SF)
As an affiliate, we earn on qualifying purchases.
What Alibaba Released on Launch Day
Alibaba’s launch announcement highlighted changes to the model’s architecture, including a combination of Gated DeltaNet layers and a system it calls Qwen Sparse Attention for long inputs. The company also described four parallel residual branches and a large lookup table for common token patterns. Alibaba said training proceeded with “no loss spikes throughout the run.” Those are vendor descriptions and claims; without public weights or detailed independent validation, they remain difficult for outsiders to check.
The company’s own Work360 benchmark covers what it describes as 26 major domains, 182 subdomains and 721 specialized fields. Alibaba reported that Qwen4 Max scored 52.69, up from 47.74 for Qwen3.8 Max. It also said average Work360 task cost fell from $4.43 to $1.67. These results are separate from Artificial Analysis’s index and cost figures, and come from an evaluation conducted by Alibaba.
Alibaba also cited a secure-coding test, where it said the share of functions passing both functionality and security checks rose from 22.0% to 31.8%. Its launch materials included examples involving manufacturing, legal review and scientific work. Those examples are company case studies, not controlled comparisons. A limited preview reportedly opened on November 11, according to The Next Web; Alibaba announced the broader launch on November 16.
“In our Work360 evaluation, the average cost per task drops from $4.43 to $1.67 — a 62% reduction.”
— Alibaba
Open Questions on Speed and Access
Latency and throughput remain unknown in the launch-day results: Artificial Analysis had no speed data for Qwen4 Max. That matters for applications where response time or the ability to handle many requests is as important as benchmark scores and token prices.
It is also not yet clear how closely Alibaba’s Work360 results will translate to independent tests or customer workloads. The company’s launch comparison chart was published as an image, and the source report said it did not reconstruct values from that chart. The Next Web, citing FourWeekMBA, reported mixed coding-test results: a 75.7 score on DeepSWE v1.1, slightly above Claude Opus 5.5’s 74.2, but 48.5 on Terminal-Bench 4.0, below Claude Opus 5.5 and Kimi K3.1. Those figures were reported from Alibaba’s comparison material and should not be confused with Artificial Analysis’s index.
Finally, Alibaba has not given a date for releasing smaller Qwen4 models as open source. Qwen4 Max itself remains API and service-based rather than downloadable, and independent verification of its architecture is limited while the weights remain closed.
Further Benchmarks and Open-Model Plans
The next useful evidence will be independent speed and latency measurements, alongside additional evaluations that make clear which model settings and task types were used. Results from organisations testing Qwen4 Max on their own workloads may also show whether the benchmark cost advantage carries over to particular applications.
Alibaba says it plans to open-source smaller models in the Qwen4 family, but has not supplied a timetable. Until those releases or additional independent testing arrive, buyers can compare the available benchmark results and API prices, but should treat claims about broader performance and case-study savings as unverified for their own use.
Key Questions
What did Artificial Analysis score Qwen4 Max?
Artificial Analysis gave Qwen4 Max a 43 on its Intelligence Index v5.0.1, tested at the xhigh reasoning setting. It scored the previous Qwen3.8 Max (0902) at 34.
Does Qwen4 Max cost a third as much as its predecessor?
Not according to the cited index-task comparison: Artificial Analysis reported $3.27 per task for Qwen4 Max and $10.02 for Qwen3.8 Max. Alibaba separately reported a fall from $4.43 to $1.67 on its Work360 benchmark. These are results from different evaluations, not a universal cost for all use.
Is Qwen4 Max open source?
No. Qwen4 Max’s weights are not available to download. Alibaba says smaller models in the Qwen4 family will be open-sourced, but has not announced when.
How does Qwen4 Max rank against other models?
It ranked ninth of 10 models on the Artificial Analysis comparison page described in the source report. Its score was above the cited comparable-model median of 41, but the leading model on that page scored 65.
Are Qwen4 Max’s speed and latency known?
No speed or latency measurements were available from Artificial Analysis on launch day, so response-time comparisons remain open.
Source: ThorstenMeyerAI.com
