Chinese AI company MiniMax releases new models it claims are competitive with the industry's best | TechCrunch
Briefly

MiniMax, backed by Alibaba and Tencent, introduced three new AI models: MiniMax-Text-01 for text, MiniMax-VL-01 for multimodal understanding, and T2A-01-HD for audio.
MiniMax-Text-01 claimed to outperform Google's Gemini 2.0 Flash on benchmarks that assess models' abilities in math problem-solving and fact-based queries, which is significant.
With a context window of 4 million tokens, MiniMax-Text-01 can analyze about 3 million words simultaneously, greatly surpassing the capabilities of GPT-4o and Llama 3.1.
While MiniMax-VL-01 is competitive against models like Claude 3.5 Sonnet, it has not consistently outperformed Gemini 2.0 Flash and others in multimodal evaluations.
Read at TechCrunch
[
|
]