By South China Morning Post | Created at 2025-01-15 23:11:44 | Updated at 2025-01-16 02:55:10
3 hours ago
Chinese artificial intelligence (AI) start-up MiniMax has launched a new series of open-source models, intensifying the competition among mainland technology firms to deliver cost-effective AI systems that can rival top offerings from US competitors.
The Shanghai-based company on Tuesday unveiled the MiniMax-01 large language model (LLM) family, which includes a general-purpose foundational model, the MiniMax-Text-01, and the multimodal MiniMax-VL-01 with visual capabilities. LLMs are the tech that powers text-generating AI products such as ChatGPT.
In benchmark tests that MiniMax posted to its official WeChat account, the new foundational language model was shown to be on par with the world’s leading AI models in evaluations that included maths problem solving, domain knowledge, the ability to follow instructions, and avoiding hallucinations or factual errors.
MiniMax’s release of an industry-leading AI system comes just weeks after Hangzhou-based rival DeepSeek took the world’s tech community by storm with its open-source V3 model in December. Growing competition in the country’s crowded and rapidly evolving AI market has pushed tech giants and start-ups alike into a race to outdo each other every few months with new releases.
Hong Kong-listed SenseTime also unveiled a new “unified large model” on Tuesday. The model combines text and image understanding with reasoning capabilities. Benchmark tests by SuperCLUE, which specialises in evaluating Chinese models, ranked the new SenseTime product as a top contender among multimodal models.
