The Talent War for Large-Scale AI Models Kicks Off
Author: Jing Xing | Editor: Yang Zhou

“Shixiang” has obtained exclusive information that MiniMax, a prominent large-scale AI model company, has recently launched a stock option incentive program worth millions of dollars to lure top talent. Based on employees' contributions to the company, MiniMax will offer incentives ranging from hundreds of thousands to several million dollars, aiming to motivate staff to boldly pursue artificial general intelligence (AGI).

Previously, ByteDance's Seed department also rolled out a stock option plan targeting technical employees in the large-scale AI model domain. More recently, another large-scale model firm, Kimi, has reported (replacing the incorrect "transmit") similar stock option incentive news.

Furthermore, “Shixiang” has exclusively learned that MiniMax's stock option incentives will be extended to employees across all departments, including technical, product, and administrative roles. As long as employees meet the talent incentive criteria, they can join the stock option incentive program. There have even been instances where interns have received incentives worth hundreds of thousands of dollars.

Public information reveals that MiniMax was established in 2022, focusing on the general artificial intelligence (AGI) sector. It boasts multiple self-developed multimodal general large-scale models, such as MiniMax M1, Hailuo-02, Speech-02, and Music-01. The company operates AI-native applications globally, including Hailuo AI, MiniMax, Xingye, and MiniMax Audio, covering sectors like emotional companionship and video generation, with a cumulative user base exceeding 157 million.

As one of the few AI unicorns capable of poaching talent from major companies, MiniMax has maintained a relatively low-key public image but has kept pace with leading firms in terms of technological capabilities. In May this year, MiniMax's TTS speech large-scale model, Speech-02, topped international evaluation rankings, outperforming giants like OpenAI and ElevenLabs, achieving state-of-the-art (SOTA) performance in key metrics such as word error rate and speaker similarity.

In June, MiniMax unveiled a series of models and tools within a week, including the open-source reasoning model MiniMax-M1, the video generation model Hailuo 02, the general agent MiniMax Agent, the video creation agent Hailuo Video Agent, and the voice design tool Voice Design.

From an industry standpoint, the development of large-scale AI models is marked by high investment, long research and development cycles, and dependence on key innovations. The phenomenon of major companies poaching talent from each other is prevalent.

An AI large-scale model practitioner told “Shixiang” that the current pace of technological iteration in the industry is extremely fast. The success or failure of companies does not depend on one or two specific product generations but rather on whether their technological positioning can keep up with the leading pack and maintain an advantage in long-term competition. In the future, scenarios like ChatGPT sweeping the entire market with a single product generation will become increasingly rare. Waging a steady and protracted war will be the key to victory.

For large-scale AI model companies, ensuring a combat-ready core team is vital. For these firms, providing competitive incentives to core talent is both a way to build defensive barriers and safeguard their talent pool, as well as a means to attract top talent with substantial financial rewards, forming a positive cycle of 'more AI talent leading to faster technological iteration, stronger product competitiveness, and further talent attraction.'

At the same time, incentivizing talent through additional stock option issuances is itself an exploration by companies to build competitive barriers. “Shixiang” has learned that this stock option incentive program will serve as MiniMax's long-term talent incentive standard and be incorporated into future incentive models. This robust and high-frequency incentive mechanism enhances employees' sense of gain and sets a new benchmark for industry competition.

Currently, the large-scale AI model market is transitioning from text-based conversations to a multimodal fusion of text, audio, images, and videos, with industry competition entering a fierce phase. The moves by multiple companies to release incentive policies indicate that AI talent competition has become paramount. Only those who can attract top talent and maintain an entrepreneurial mindset can hold the high ground in the competition.

This article is an original piece from “Shixiang”
Unauthorized reproduction or mirroring is prohibited.