Following the release of the initial wave of third-party validation data, JPMorgan has adopted a more positive view on the M3 model from MINIMAX-WP (ASX: 00100).
In its initial assessment on June 1st, the firm noted that M3's premium pricing was a strong signal of MiniMax's ambition to achieve state-of-the-art (SOTA) performance, though it lacked independent user feedback, third-party benchmark validation, and real-world token consumption data at the time.
Subsequently, platforms including Artificial Analysis, Code Arena, and OpenRouter have begun providing more evidence. M3 has now ranked among the top-tier Chinese frontier models on Artificial Analysis. On the Code Arena WebDev leaderboard, it ranks 7th by model and 4th by laboratory, with scores approaching those of Zhipu's GLM-5.1. Shortly after its launch, daily token usage on OpenRouter surpassed 500 billion.
However, this new positive data has not fully alleviated market concerns regarding its sustained pricing power. The firm believes the next crucial validation lies in user retention. If usage on OpenRouter remains robust after the 50% discount period concludes, and if M3 continues to gain greater adoption within coding tools, MiniMax's high-end model strategy will become more compelling, strengthening its narrative around Annual Recurring Revenue (ARR) quality.
Conversely, if token usage declines significantly post-discount or feedback from coding tools is mixed, the market may continue to question whether M3's quality advantage is sufficient to justify its premium pricing against competitors such as DeepSeek.
Comments