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

TL;DR

Alibaba launched Qwen4 Max on Nov. 16, 2026, describing a 2.4-trillion-parameter model with a one-million-token context window. Artificial Analysis scored it 43, up from 34 for its predecessor, and estimated a $3.27 cost per index task, compared with $10.02 for the previous model; speed data was not available on launch day.

Alibaba launched Qwen4 Max on Nov. 16, 2026, and the benchmarking company Artificial Analysis published its first results that day: the new model scored 43 on its Intelligence Index, compared with 34 for its predecessor, while an index task was estimated to cost $3.27 rather than $10.02. The score is above the 41 median cited for similarly priced models, but Qwen4 Max ranked ninth among the 10 models listed on its Artificial Analysis page; speed measurements were not yet available.

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

Alibaba describes Qwen4 Max as the first model in its Qwen4 family, with 2.4 trillion backbone parameters, 71 billion activated per token and a one-million-token context window. The company says it can process text, images and video. Its launch materials call it a “24/7 AI teammate” and describe changes to attention and model architecture. Those specifications and descriptions come from Alibaba; the model weights are closed, so outside researchers cannot inspect them directly.

Artificial Analysis’s launch-day comparison shows a nine-point increase over Qwen3.8 Max (0902), which scored 34. On the same page, Qwen4 Max scored below the listed leading model’s 65. Artificial Analysis estimated a cost of $3.27 per Intelligence Index task for Qwen4 Max, compared with $10.02 for its predecessor. Alibaba’s listed API prices are $1.50 per million input tokens and $4.50 per million output tokens. The index-task estimate and token prices describe different measures of cost.

Alibaba also reports gains on its in-house Work360 benchmark: Qwen4 Max scored 52.69, up from 47.74 for Qwen3.8 Max, while the company says average cost per task fell from $4.43 to $1.67. Work360 is Alibaba’s own evaluation, and its results have not been independently verified in the supplied material. Artificial Analysis had no speed or latency figures for Qwen4 Max at publication.

At a glance
reportWhen: Launched and independently benchmarked…
The developmentArtificial Analysis published its first Qwen4 Max benchmark results on the same day Alibaba launched the model, showing improved scores and lower measured task costs than its predecessor.
Bar chart titled Ninth of ten on its own page, nine points above its predecessor. Artificial Analysis Intelligence Index for the 10 models on the Qwen4 Max page: Claude Fable 5.2 (max) 65, Gemini 4 Carbon (high) 60, Claude Opus 5.5 (max) 59, GPT-6 Astra (max) 57, GPT-6.1 Sol (max) 55, Muse Wonder Full (max) 50, Kimi K3.1 Pro (open weights) 46, Muse Wonder (max) 44, Qwen4 Max 43 (highlighted), Grok 4.7 (xhigh) 35. Below a dashed line: the predecessor Qwen3.8 Max (0902) 34
Qwen4 Max is ninth of the ten models on its Artificial Analysis page and nine points above its predecessor. Source: Artificial Analysis, model pages for Qwen4 Max and Qwen3.8 Max (0902), retrieved November 16, 2026. Chart: Thorsten Meyer AI.

Lower Costs, With Performance Limits

The results give businesses a measurable reason to consider Qwen4 Max: on Artificial Analysis’s estimate, an index task costs about one-third as much as the same measure for Qwen3.8 Max, alongside a higher index score. For teams handling many model calls, the cost difference could affect operating budgets. The benchmark does not establish what a particular business will pay for its own workload; actual API bills depend on input and output token use and the selected service.

The score also sets a limit on the launch claim. A 43 is an improvement over the previous Qwen flagship and sits just above the cited median, but it remains below the top score of 65 on the same page. That makes the evidence relevant to buyers weighing price against benchmark performance, while leaving practical questions about response speed and task-specific quality unanswered. Alibaba’s Work360 figures suggest further savings, but they are vendor-reported results, not an independent replication.

Products Worth Considering

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Anker SOLIX C1000 Gen 2 Portable Power Station with 200W Solar Panel

Anker SOLIX C1000 Gen 2 Portable Power Station with 200W Solar Panel

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

As an affiliate, we earn on qualifying purchases.

A Closed Model in the Qwen4 Family

Qwen4 Max is available through Alibaba’s QwenCloud API, a subscription called Token Plan, and the company’s QwenWork, Qoder and QwenStudio apps, according to Alibaba’s launch materials. The model is not available for download. Alibaba says smaller Qwen4 models will be open-sourced, but the supplied source material gives no release date. The Next Web reported that a limited preview began on Nov. 11, before the public launch.

Alibaba’s launch post describes its own Work360 evaluation as covering 26 Tier-1 domains, 182 Tier-2 subdomains and 721 Tier-3 fields. The company also reports lower token use on selected tests and a higher pass rate on a secure-coding evaluation. Those figures are claims from Alibaba’s benchmark reporting. The launch material includes other demonstrations and partner case studies, but they do not establish how the model performs across typical customer workloads.

“Your new teammate has arrived.”

— Alibaba’s Qwen team

Speed and Real-World Costs

Speed and latency remain unmeasured in the launch-day Artificial Analysis results, so buyers cannot use the supplied figures to compare response time. It is also unclear how the index score translates to particular tasks, workloads or service settings. The benchmark ranking describes the models listed on the Qwen4 Max page, not every available AI model.

Alibaba’s Work360 results, token-use reductions and case studies are vendor claims in the material available here. The source does not provide an independent reproduction of those results. The weights are closed, limiting outside inspection of the model itself. Alibaba has said smaller Qwen4 models will be open-sourced, but a timetable is not specified.

More Testing and Model Releases

Further independent benchmarks, including speed and latency measurements, would help clarify how Qwen4 Max compares beyond the launch-day Intelligence Index. Buyers can also compare their own tasks and token usage against Alibaba’s API prices; the supplied results do not predict the cost of every workload.

The next family development identified in Alibaba’s announcement is the release of smaller Qwen4 models as open-source models. No release date is given in the source material. Until those models appear and outside evaluations expand, the clearest confirmed comparison remains the initial one: Qwen4 Max scored higher and had a lower estimated index-task cost than its predecessor, while trailing the page’s leading model.

Key Questions

What did Artificial Analysis score Qwen4 Max?

Artificial Analysis gave it 43 on the Intelligence Index. The source material says Qwen3.8 Max (0902) scored 34 and the top model listed on the Qwen4 Max page scored 65.

How does its estimated task cost compare with its predecessor?

Artificial Analysis estimated $3.27 per index task for Qwen4 Max, compared with $10.02 for Qwen3.8 Max. This is an index-task estimate, separate from Alibaba’s per-token API prices.

Is Qwen4 Max open source?

No. The source material says Qwen4 Max’s weights are closed and that it is offered through Alibaba services. Alibaba says smaller Qwen4 models will be open-sourced, but has not supplied a date here.

Are speed and latency results available?

No speed or latency data was available from Artificial Analysis on launch day, according to the supplied source. The available comparison therefore does not establish how quickly Qwen4 Max responds.

Source: Thorsten Meyer AI

You May Also Like

Claude Fable 5.2 At #1: How Effort Levels Change Your Cost Per Task

Artificial Analysis ranks Fable 5.2 first at Max effort, where an index task costs $26.54; High scores nearly as well for $9.08.

Beam: Reflection’s 501B Open-weight Model

Reflection says Beam has 501B total parameters and 23B active, with a focus on coding and agentic tasks. Weights and evaluations are still pending.

A TV Box Botnet, Factory Malware, And Google’s Takedown

Google says it disrupted the XCPE botnet, linked to malware installed in TV-box firmware. The backdoor may remain on affected devices.

What Happens When AI Leaders Need More Memory?

Sam Altman met Samsung executives in South Korea amid rising demand for memory chips. No new supply agreement was announced.