Signal

Qwen4-Max scores 43 on the Artificial Analysis Intelligence Index, beating Qwen3.8 Max at ~67% lower cost per task but trailing Kimi K3.1 Pro at ~11% lower cost

First reported by X ·

The signal ●●●○ Compiled by AI from X and Techmeme
Why you might care

LLM task costs decrease by up to 67% for comparable intelligence.

What happened

Alibaba has released Qwen4 Max, a new large language model that scores 43 on the Artificial Analysis Intelligence Index. This represents a 9-point improvement over its predecessor, Qwen3.8 Max. Qwen4 Max offers a cost per task that is approximately 67% lower than Qwen3.8 Max, though it trails Kimi K3.1 Pro by 3 points on the index, with Kimi K3.1 Pro costing about 11% more per task. The Qwen4 Max model utilizes a high reasoning effort setting for its evaluation. This benchmarking data was provided by Artificial Analysis, an independent entity that analyzes AI science and technology. The evaluation also highlights the increasing importance of faster token generation for multi-step agentic workflows.

What it means

The release of Qwen4 Max signifies a continued trend of rapid iteration and performance gains in the LLM market, particularly from major tech players like Alibaba. While Qwen4 Max demonstrates significant improvement and cost-efficiency over its previous version, its performance relative to Kimi K3.1 Pro suggests an ongoing competitive landscape where incremental gains in intelligence are met with varying cost structures. This dynamic directly impacts developers and businesses selecting LLMs, forcing a trade-off between peak performance and operational expenditure.

The data underscores that developers must carefully consider both benchmark scores and cost-per-task metrics when integrating LLMs into applications. As models like Qwen4 Max become more affordable for similar or slightly lower intelligence compared to top-tier options, the economic viability of advanced AI capabilities expands. The mention of faster token generation also points to future developments focusing on efficiency for complex, agent-like tasks, which could further redefine cost-performance benchmarks.

AI-written summary. May contain errors.

Chips