Beating AI News Flash: Independent evaluation agency Artificial Analysis has published the results for the official version of Qwen4-Max. At the highest xhigh reasoning tier, the model achieved an overall intelligence index score of 43 points, an increase of 9 points over the previous generation Qwen3.8-Max's 34 points, and 3 points short of Kimi K3.1 Pro's 46 points.
Qwen4-Max surpassed its predecessor in all 9 evaluations. On the high-difficulty knowledge reasoning test HLE-Diamond, accuracy improved from 24% to 45%, and the enterprise software automation test improved by 17 percentage points. In AA-Briefcase, which simulates real-world business document processing, Qwen4-Max scored 1686 Elo, slightly above Kimi K3.1 Pro's 1674.
Coding capability remains a weak spot. On the Terminal-Bench 4.0 test, Qwen4-Max scored 48%, clearly trailing Kimi K3.1 Pro's 67%. On FrontierCode 2.0, which assesses real-world open-source project development capability, both scored 26%, behind Tencent Hy4's 31%.
On cost, Qwen4-Max's average per-task cost is $3.27, down from $10.02 for the previous generation, a reduction of about 67%, and also about 11% cheaper than Kimi K3.1 Pro's $3.68.
In addition to API price cuts, the model itself also reduced Token consumption. Qwen4-Max outputs an average of about 105,000 Tokens per task, a 39% decrease from the previous generation's 171,000. However, Kimi K3.1 Pro requires only about 72,000 Tokens.
