Chinese AI startup MiniMax has announced the release of its latest large language model, MiniMax-M1, setting a bold challenge to leading competitors — including China’s own DeepSeek. The Shanghai-based company claims the M1 model not only surpasses DeepSeek R1-0528 in several reasoning benchmarks but also achieves that performance with significantly less computing demand.
Performance Highlights
According to MiniMax, the M1 model supports a context length of one million tokens, a substantial leap over DeepSeek’s R1, which supports just 128,000. Context length — the amount of data the model can process simultaneously — is a critical metric in determining an AI’s ability to reason over large bodies of information.
M1 also reportedly uses only 30% of the computing resources required by DeepSeek under comparable circumstances, offering efficiency as well as scale. The company emphasized that these results were achieved using large-scale reinforcement learning across 512 Nvidia H800 GPUs, with total compute rental valued at $534,700.
“M1 is designed for productivity and complex reasoning. It’s the most efficient closed-source Chinese LLM today,” MiniMax stated.
While Bloomberg has not independently verified the results, the release is already stirring conversations within China’s fast-evolving AI landscape.
Backed by Tech Giants
MiniMax is part of the so-called “Little Dragons” — a group of elite Chinese AI startups that includes some of the most well-funded players in the space. Backed by internet giants Tencent Holdings Ltd. and Alibaba Group Holding Ltd., the startup has raised billions over the last year. These companies have been racing to develop models competitive with OpenAI, Google DeepMind, and other global leaders.
However, the rapid ascent of DeepSeek last year shifted strategic priorities across the AI sector in China. Several of the Little Dragons were forced to pivot away from fundamental research to focus on applications. MiniMax’s latest announcement signals a return to core innovation and model development.
Applications and Next Steps
MiniMax’s current product lineup includes a video generation platform and an AI companion app, with the M1 model expected to power new and upcoming offerings. The company stated that more technical details and capabilities will be released in the coming days.
Why It Matters
The release of M1 reflects a growing ambition among Chinese firms to compete globally on AI reasoning and efficiency, beyond chat applications or text summarization. If MiniMax’s claims are substantiated, the M1 model could set a new benchmark for what’s possible in domestic AI — and challenge the current dominance of DeepSeek in the Chinese market.
As the AI race intensifies in both China and globally, MiniMax’s M1 could mark a turning point, both in innovation and in the balance of AI power in the East.