Chinese companies are continually releasing AI models that are competitive with those developed by U.S.-based AI companies like OpenAI.
This week, MiniMax, a startup backed by Alibaba and Tencent, introduced three new models: MiniMax-Text-01, MiniMax-VL-01, and T2A-01-HD. MiniMax-Text-01 is a text-only model, MiniMax-VL-01 understands images and text, and T2A-01-HD generates speech.
MiniMax claims that MiniMax-Text-01 outperforms Google’s Gemini 2.0 Flash on benchmarks like MATH and SimpleQA. MiniMax-VL-01 competes with Anthropic’s Claude 3.5 Sonnet on tasks requiring multimodal understanding like ChartQA.
MiniMax-Text-01 has an extensive context window, significantly larger than other models like GPT-4o and Llama 3.1’s.
T2A-01-HD is an audio generator optimized for speech in multiple languages.
MiniMax’s models are available on GitHub and Hugging Face, with T2A-01-HD exclusively accessible through MiniMax’s API and Hailuo AI platform.
MiniMax was founded by former SenseTime employees and has released controversial products like Talkie featuring AI avatars of public figures.
MiniMax’s new models launch amidst proposed export rules by the Biden Administration aimed at restricting advanced AI technologies for Chinese ventures.
