MiniMax Updates Performance and Price Simultaneously with New Model

Chinese artificial intelligence startup MiniMax announced a new language model 'M3.1 Flash.' The model achieved a score of 42 on the 'Artificial Analysis Intelligence Index (AAII),' which comprehensively evaluates AI performance, representing a 13-point improvement from the previous generation model. What is particularly noteworthy is its price point, offered at half the level of its predecessor.

In China's AI industry, a trend has continued in recent years where model performance improvements and price reductions occur simultaneously. There has been a trajectory where the use of large language models (LLMs), originally considered high-cost, has become significantly cheaper due to aggressive cost competition among Chinese companies, and M3.1 Flash can be seen as an extension of this trend. Such price competition has become a factor in the widespread adoption of AI services.

On the 'DeepSWE v1.1' metric, which measures coding (programming) performance, M3.1 Flash recorded a score of 73.7%. At a cost of less than $0.30 per task (approximately 45 yen), both performance and price are reported to be equivalent to DeepSeek's 'V4.1 Flash 1017.' Coding performance that supports automated program generation and modification is one of the high-demand areas in enterprise AI adoption.

The significance of these figures extends beyond simple benchmark comparisons. The 13-point improvement in AAII score demonstrates a major leap from the previous generation, and considering the difficulty of model development that simultaneously achieves lower pricing and performance enhancement, this result is worthy of attention. The fact that performance and price are now on par with competitor DeepSeek is also an example of how intense competition in the Chinese AI market has become.

From an industry-wide perspective, the moves by Chinese AI companies to reduce prices while maintaining and improving performance could impact the global AI cost structure. In particular, when enterprises and developers integrate AI into practical products and services, the per-task cost is a factor directly linked to decision-making. The emergence of low-cost, high-performance models like M3.1 Flash has the potential to further lower the barrier to entry for AI-powered services.

The key point to watch going forward is how far this price competition will continue. At the present time, the 'bottom' of pricing is not yet visible, and whether MiniMax and other Chinese AI players can further compress costs while improving performance will become the next focus of the industry. Additionally, it is necessary to continue monitoring how these developments will influence the pricing strategies of AI companies outside China.