Xiaomi Enters AI Race with MiMo-7B, Challenging OpenAI and Alibaba

China AI rising: Xiaomi releases new MiMo-7B models as DeepSeek upgrades its Prover math AI - SiliconANGLE

China AI rising: Xiaomi releases new MiMo-7B models as DeepSeek upgrades its Prover math AI Xiaomi Corp. today released MiMo-7B, a new family of reasoning models that it claims can outperform OpenAI's o1-mini at some tasks. The algorithm series is available under an open-source license. Its launch coincides with DeepSeek's release of an update to Prover, a competing open-source reasoning model. The latter algorithm has a narrower focus than MiMo-7B: it's designed to help mathematicians prove theorems. The algorithms in Xiaomi's MiMo-7B series have about seven billion parameters. There's a base model, as well as enhanced versions of that model that offer increased output quality. Xiaomi developed the enhanced versions using two machine learning techniques called supervised fine-tuning and reinforcement learning. Both methods improve AI models by providing them with additional training data. The datasets used in supervised fine-tuning include explainers that help guide the AI training workflow, while reinforcement learning doesn't use such explainers. Xiaomi has developed three enhanced versions of the MiMo-7B base model. It fined-tuned one version using supervised fine-tuning, another with reinforcement learning and a third using both methods. According to the company, that third model is better at OpenAI's o1-mini at generating code and solving math problems. The base MiMo-7B model is less capable than the fine-tuned versions, but can still outdo significantly larger algorithms. "Our RL experiments from MiMo-7B-Base show that our model possesses extraordinary reasoning potential, even surpassing much larger 32B models," Xiaomi researchers detailed on GitHub. The MiMo-7B series is not the only new entry into the open-source AI ecosystem that debuted today. DeepSeek quietly released an enhanced version of Prover, a reasoning model optimized to prove mathematical theorems that it first debuted last year. Prover-V2, as the upgraded model is called, promises to provide "state-of-the-art performance in neural theorem proving." DeepSeek trained Prover-V2 through a multi-step process. The company started by assembling a collection of theorems for which proofs are already available. In the next step, DeepSeek used two language models to create a step-by-step explanation of how mathematicians arrived at each proof. The company subsequently entered these AI-generation explanations into Prover V2 to teach the model how to generate its own proofs. "This process enables us to integrate both informal and formal mathematical reasoning into a unified model," DeepSeek researchers explained.

Xiaomi joins China AI game with maiden DeepSeek-like model

Xiaomi has launched its open-source AI model. It joins other Chinese tech giants in this field. The model is named MiMo. It mimics human thinking. Xiaomi claims MiMo outperforms OpenAI and Alibaba platforms in tests. This launch follows Alibaba's new model release. It also comes after DeepSeek's R1 success. Xiaomi aims to develop human-level AI.Xiaomi Corp. unveiled its own open-source AI model, joining the growing ranks of Chinese tech leaders hoping to make a splash in a burgeoning field endorsed by Beijing. The smartphone and EV maker publicized the MiMo reasoning model, which like DeepSeek's R1 mimics the way humans think through problems. Xiaomi printed stats on WeChat showing it surpassed OpenAI's o1-mini and Alibaba Group Holding Ltd.'s Qwen platforms in benchmark tests, which Bloomberg hasn't independently verified. Xiaomi debuted MiMo a day after Alibaba unveiled the latest version of its own flagship model, amplifying a race between China's tech players that erupted after DeepSeek's R1 stunned the industry with a low-cost yet capable platform. Better-known for smartphones and most recently a line-up of EVs, Xiaomi declared on Wednesday it too was now intent on developing artificial general intelligence, or AI with human-level intellect. Its shares rose more than 5% in Hong Kong. Chinese AI stocks also climbed Wednesday after broad state media coverage of President Xi Jinping's visit to an incubator and accelerator for AI tech firms, underscoring Beijing's support for homegrown tech that can compete with the US. "MiMo is the first taste of our newly established core AI model team," Xiaomi said in a WeChat post. "While 2025 may seem like a late stage to take up the dream of large models, we believe AGI is a long-term endeavor." Xiaomi executives have reportedly talked about investing in AI in the past, though MiMo is the first real product to emerge. Its foray marks the second big project in as many years for a company best-known for making affordable smartphones and appliances from robot vacuums to rice cookers. Billionaire founder Lei Jun spearheaded a $10 billion drive into the crowded EV arena in 2024, calling it his final startup hurrah. Bloomberg News reported it was forced to push back the debut of its first sport utility vehicle following a fatal accident involving its marquee electric car, a blow to its ambitions to put pressure on Tesla Inc. and BYD Co. Its stock has plunged some 15% since that March episode. In a posting on Weibo after Bloomberg's report was published, the company said the timeline for YU7's release remains unchanged and is still set for June or July.

Xiaomi joins AI race with DeepSeek-like debut model MiMo; Shares surge By Investing.com

Investing.com-- Xiaomi Corp (HK:1810) on Wednesday debuted its first open-source artificial intelligence large language model, MiMo, joining an increasingly crowded stable of technology giants seeking to gain a foothold in the fast-growing industry. The launch of the new model saw Xiaomi's shares surge nearly 5% to HK$50.0 in Hong Kong trade, helping underpin small gains on the Hang Seng. Xiaomi (OTC:XIACF) unveiled the model in an announcement, stating that the model is designed specifically for reasoning tasks, making it similar in function to DeepSeek's R1. The Chinese tech giant claimed that MiMo outperformed OpenAI's o1-mini and Alibaba's QwQ-32B-Preview in mathematical reasoning and coding benchmarks. Xiaomi said DeepSeek-R1 had triggered a wave of reinforcement learning development in the AI industry. But with the same reinforcement learning training data, Xiaomi claimed that MiMo "has a significantly leading reinforcement learning potential in the field of mathematics & code." The move sees Xiaomi joining a cadre of its Chinese tech peers in attempting to gain more market share in the increasingly competitive AI space. So far, Alibaba (NYSE:BABA), DeepSeek, and TikTok-owner ByteDance lead Chinese markets by user numbers for their AI tools. China's new wave of AI development was sparked by DeepSeek's R1 model earlier this year. The model highlighted China's AI development capabilities despite a lack of access to advanced hardware, amid strict U.S. controls on tech exports to China. While Xiaomi does offer AI services through its Xiaomi HyperAI tool, this is largely through a partnership with Google (NASDAQ:GOOGL) and uses Google's Gemini AI model. The Chinese electronics giant has been steadily attempting to diversify beyond its core smartphones and devices business, having recently made a successful foray into electric vehicles with its SU7 lineup.