Chinese tech startup, MiniMax, recently launched three AI models it claims to challenge the prowess of industry leaders. The introduction of MiniMax-Text-01, MiniMax-VL-01, and T2A-01-HD models aim to reform the global pattern in AI technology.
Endowed with an impressive 456 billion parameters, the text-only model, MiniMax-Text-01 outperforms Google’s Gemini 2.0 Flash on certain metrics. Moreover, MiniMax also claims that MiniMax-VL-01, an image-text combo model, equally performs on par with Claude 3.5 Sonnet by Anthropic on evaluations requiring multicodal comprehension.
The new models also include T2A-01-HD, an audio generator primarily centered on speech production, capable of generating synthetic voice with adjustable cadence in multiple languages.
Unlike other new models, T2A-01-HD is exclusively available only through MiniMax’s API and Hailuo AI platform. However, despite being ‘openly’ available, MiniMax-Text-01 and MiniMax-VL-01 maintain certain restrictions under the restrictive license from MiniMax.
These models by MiniMax, backed by Alibaba and Tencent, encompass their effort to rival the capabilities of AI models by OpenAI and similar US-based AI companies.
Original source: Read the full article on TechCrunch
