初创公司 Mistral 的 AI 模型某些方面可与 GPT-4 媲美

软餐获悉,由谷歌前员工创办的初创公司 Mistral AI 引发关注。31 岁的 Mensch 与两位朋友创办了这家公司,Mistral 在本周一发布了一款名为 Mistral Large 的新 AI 模型,该模型在执行一些推理任务时,可以与 OpenAI 迄今为止最先进的语言模型 GPT-4 和谷歌的新模型 Gemini Ultra 相媲美。而 Mistral Large 的训练成本不到 2,000 万欧元(约合 2,200 万美元),相比之下,GPT-4 的训练成本为 5,000 万至 1 亿美元。Mistral 公司认为,相比巨头,它会找到更高效的方法来构建和部署 AI 系统。该公司已筹集到 5 亿美元融资,投资者包括 Andreessen Horowitz,微软、Salesforce 和英伟达也持有 Mistral 的少量股份。


Mistral AI, a startup founded by former Google employees, has garnered attention. The 31-year-old Mensch, along with two friends, established the company. Mistral unveiled a new AI model called Mistral Large on Monday, which is on par with OpenAI’s most advanced language model, GPT-4, and Google’s new model, Gemini Ultra, in performing certain inference tasks. However, the training cost for Mistral Large is less than 20 million euros (approximately 22 million dollars), while the training cost for the GPT-4 model ranges from 50 million to 100 million dollars. Mistral believes that, compared to industry giants, it will find more efficient ways to build and deploy AI systems. The company has raised $500 million in funding, with investors including Andreessen Horowitz, as well as minority stakes held by Microsoft, Salesforce, and NVIDIA.

初创公司Mistral的AI模型某些方面可与GPT-4媲美
(0)
无忌的头像无忌

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注