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

🔍 Read the full analysis: The Cost And Capability Trade-Off Behind Qwen4 Max’s Latest Results 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 gave Alibaba’s Qwen4 Max an Intelligence Index score of 43, up from 34 for its predecessor, and reported a lower cost per index task. The model ranked ninth among 10 on its comparison page, and independent speed data was unavailable at launch. Alibaba’s additional benchmark and business results are company-reported, while Qwen4 Max’s weights remain closed.

Alibaba’s new Qwen4 Max scored 43 on Artificial Analysis’s Intelligence Index, nine points above its predecessor, while costing less per index task, according to the benchmarking company’s results published Nov. 16. The model’s score was above the comparison group’s stated median of 41, but it placed ninth among 10 models on its Artificial Analysis page, leaving a mixed picture of improved capability and relative standing.

Artificial Analysis reported that a Qwen4 Max index task cost $3.27, compared with $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. These figures describe different measures: the benchmark task cost is not the same as a fixed per-token price, because task cost depends on how many tokens a run uses.

The index combines nine evaluations spanning areas including agentic coding, office automation, knowledge and physics. Artificial Analysis tested Qwen4 Max at its xhigh reasoning setting. Alibaba’s API default is reported as high, so the benchmark result does not establish what users will score with default settings. On the same comparison page, the leading model scored 65; the highest listed Chinese model was Kimi K3.1 Pro at 46.

Alibaba describes Qwen4 Max as a 2.4-trillion-parameter backbone with 71 billion parameters activated per token, plus 328 billion N-gram embedding parameters. The company says it supports text, images and video and has a one-million-token context window. The model is available through Alibaba’s API and named company products, but its weights are not available to download.

At a glance
reportWhen: Published Nov. 16, 2026; launch-day tes…
The developmentArtificial Analysis published its first Qwen4 Max results on Nov. 16, reporting a higher score than the previous Qwen flagship and lower cost per benchmark task.

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

Lower Costs, But Not a Top Rank

The results put the model’s trade-off into practical terms: compared with its predecessor, it delivered a higher score at a lower measured task cost. For organizations evaluating models for repeated workloads, cost per task can matter alongside headline capability. But the index ranking cautions against reading the improvement as evidence that Qwen4 Max leads the wider field: it scored above the stated median while remaining ninth on the page’s 10-model comparison.

The cost figures also do not settle what a buyer would pay for a particular workload. Actual API bills depend on token use, reasoning settings and task design. Alibaba says its model uses fewer tokens on some of its own evaluations, but those figures have not been independently confirmed in the supplied material. Without comparable speed and latency results, the available data cannot show whether lower task cost comes with trade-offs in response time or throughput.

Products Worth Considering

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

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

As an affiliate, we earn on qualifying purchases.

How Alibaba Frames the Launch

Alibaba introduced Qwen4 Max on Nov. 16 as the first model in the Qwen4 family, calling it a “24/7 AI teammate.” The company describes a sparse mixture-of-experts design and changes to attention and residual layers. It also says training proceeded without loss spikes. These are company descriptions; because the model weights are closed, outside researchers cannot inspect the architecture or independently reproduce the training claims from the weights.

Alibaba’s Work360 benchmark gives Qwen4 Max a score of 52.69, versus 47.74 for Qwen3.8 Max, according to the company. Alibaba also reports that average Work360 task cost fell from $4.43 to $1.67. It cites gains on a secure-coding test and case studies involving manufacturing and other professional tasks. Those are vendor-reported results, not independent measurements or controlled comparisons in the material available here.

The launch article also presents results for coding tests, including a reported DeepSWE v1.1 score of 75.7. The supplied account attributes those figures to reporting by The Next Web, citing FourWeekMBA, and says Alibaba’s broader comparison chart was published as an image rather than in machine-readable figures. Those secondary reports should not be treated as equivalent to the Artificial Analysis index results.

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

— Alibaba

What the Launch Data Cannot Show

Speed and latency data were not available from Artificial Analysis on launch day, according to the source material. It is therefore not yet possible to compare response time or throughput with rival models from the cited independent results. The benchmark score also reflects testing at xhigh reasoning effort, not necessarily the experience under Alibaba’s reported high default.

Independent confirmation is also limited. Alibaba’s Work360 results, token-use reductions, security benchmark and customer examples are company claims; the supplied material does not provide independent replication or enough test detail to evaluate all comparisons. The weights are closed, preventing outside inspection of the model. Alibaba said smaller Qwen4 models would be open-sourced, but no release date was given.

The reported scores from other coding benchmarks rely on secondary coverage of Alibaba’s chart, and the supplied account does not independently verify them. It remains unclear how Qwen4 Max will perform across a broader range of real-world workloads, what costs users will see for their own tasks, and whether later testing will change its relative ranking.

Further Tests and Open Releases

Artificial Analysis or other independent evaluators may publish additional speed, latency and task-level results as testing continues. Those measurements would help readers distinguish the launch-day index score from practical performance under everyday API settings. Customers can also compare their own workload costs against Alibaba’s published token prices, while keeping vendor benchmark results separate from independently run tests.

Alibaba has said that smaller Qwen4 models will be open-sourced, but has not announced a date in the supplied material. Until then, Qwen4 Max remains a closed-weights offering available through Alibaba’s services. Any future weight release, independent replication of company benchmarks, or updated model comparison could clarify the balance between capability, cost and access.

Key Questions

How did Qwen4 Max perform in the independent index?

Artificial Analysis gave it an Intelligence Index score of 43, nine points above Qwen3.8 Max (0902). It ranked ninth among 10 models on the Qwen4 Max comparison page and was above the stated comparison median of 41.

How much did an index task cost?

Artificial Analysis reported a cost of $3.27 per index task for Qwen4 Max, compared with $10.02 for Qwen3.8 Max (0902). The figure is a benchmark-task cost, distinct from Alibaba’s per-token API prices.

Are Alibaba’s Work360 results independently verified?

Not in the supplied material. Alibaba reported a Work360 score increase and lower average task cost, but these are vendor-reported results from its own benchmark, rather than findings independently confirmed here.

Can developers download Qwen4 Max’s weights?

No. The source material says Qwen4 Max is closed-weight and available through Alibaba services. Alibaba has said smaller Qwen4 models will be open-sourced, but has not provided a release date.

What performance information is still missing?

Artificial Analysis had no speed or latency data for the model on launch day. Independent results across additional workloads and confirmation of Alibaba’s in-house benchmark claims are also not established in the supplied reporting.

Source: ThorstenMeyerAI.com

You May Also Like

How To Edit AI Weights At Home Without Legal Consequences

A home experiment made a local AI answer 10 of 10 legal benchmark tasks, but its legal score stayed at zero and citations were often wrong.

16,000 RIAs Heard Schwab’s Claude Pitch; Small Firms Face A Seat-Count Hurdle

A headline-only report says Schwab promoted Claude to its RIA network, but a reported 20-seat minimum and claims about small firms remain unverified.

Anthropic’s Reported Gas Generator Deal: What To Know Before Drawing Conclusions

Piper Sandler confirms a $3.4 billion sale to an unnamed AI lab. The Financial Times reports Anthropic is the buyer; key deal details remain unconfirmed.

The 7 Best Thunderbolt Docking Stations To Shop In 2027

A buyer’s guide compares seven Thunderbolt docks, including Anker, Lenovo, Dell and Plugable models, by ports, charging, displays and compatibility.