AI News
All stories
Reported

Alibaba's Qwen4-Max ships, scoring 43 on the Intelligence Index at lower cost

Alibaba's Qwen4-Max is now shipping, scoring 43 on the Intelligence Index at its highest reasoning setting, up from 34 at preview.

Models · · 1 min read

Alibaba's Qwen4-Max is now a shipping model. At its highest reasoning setting, it scores 43 on the Intelligence Index, up from 34 when it was in preview. The model also generated about 39% fewer output tokens, which accounts for part of the cost drop. A lower per-token price and fewer tokens both push the bill down.

Alibaba says the model has 2.4T parameters in total but activates 71B of them per token, down 25% from the number active before. It accepts images and text and has a 1M-token context window. Smaller open-weight Qwen4 models are still to come.

The benchmark placements are mixed. Arena ranks Qwen4-Max #14 overall on Agent Arena, where it is the top Chinese model, and #6 on Code Arena WebDev with a score of 1750, 74 points above its predecessor. Vals ranks it #11 overall, with large gains in coding and finance. On the vision side, Roboflow calls the model underwhelming on its VLM benchmark, with Google, OpenAI and Anthropic now far ahead. Roboflow does note that at high reasoning effort the model is about 4.5x faster and about 68% cheaper per image than Qwen3.8-Max.

On HLAB, Qwen4-Max gets 95.7% of checklist items right but passes only 21.3% of tasks.

What changed

Alibaba's Qwen4-Max is now shipping, scoring 43 on the Intelligence Index at its highest reasoning setting, up from 34 at preview.

Type: New capability · Applies to: Qwen4-Max

Why it matters

A lower cost per task at a higher benchmark score makes Qwen4-Max a price-competitive option for agent workloads. The legal benchmark shows that checklist accuracy can overstate task completion.

What you can do

Test Qwen4-Max on your own task suite, measuring pass rate and cost per completed task, before switching any workload.

Original source

We have no link to the original source for this story yet.

Where we saw it

1 source

Latent Space

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works