Haitong International has released a research report initiating coverage on KNOWLEDGE ATLAS (02513) and MINIMAX-WP (00100), assigning both an "Outperform" rating. The firm set target prices of HK$860 for KNOWLEDGE ATLAS and HK$1,400 for MINIMAX-WP. The target price for KNOWLEDGE ATLAS implies a projected 2028 price-to-sales ratio of 70 times.

The competitive landscape in the technology model sector is evolving rapidly, with some observers suggesting the pace may be too fast, as evidenced by frequent quarterly shifts in model rankings. Nevertheless, the report notes that current market participants are on a much more solid footing compared to the initial wave of LLM startups that emerged following the advent of ChatGPT.

Haitong International believes that a more predictable competitive environment will enhance the chances for MINIMAX and KNOWLEDGE ATLAS to continue thriving and potentially become part of the "endgame" in the artificial intelligence era. The firm advises investors to look beyond quarterly ranking changes and instead focus on the foundational factors behind the current success of KNOWLEDGE ATLAS and MINIMAX, including their pace of innovation and the ongoing development of practical applications.

Both companies have demonstrated an ability to iterate their model versions at an accelerated pace, with the application of various model capabilities becoming increasingly evident. This is reflected in KNOWLEDGE ATLAS's recent price increase for its coding products and MINIMAX's sustained leading position on the OpenRouter usage rankings.

While valuation is important, the report suggests that for MINIMAX and KNOWLEDGE ATLAS, both companies exhibit characteristics more akin to venture investments at their current stage, despite being publicly listed. Investors are advised not to be deterred but to value the opportunity to gain exposure to these pure-play top-tier model companies through public markets.

Also noteworthy is the projected rise in AI adoption rates by 2026, such as increased usage of agents and multimodal applications, which is expected to boost overall market sentiment for AI and drive token consumption for underlying models.