Upload README.md with huggingface_hub

README.md

@@ -5,7 +5,8 @@ language:
 - en
 ---
 # Introduction
-新模型介绍，待定....
+On May 25, MiniMax officially released and open-sourced MiniCPM5-1B, a new-generation edge-side text foundational large language model. With a parameter scale of 1 billion, the model achieved a score of 17.9 on the AA-Index benchmark, outperforming all open-source foundational models with fewer than 4 billion parameters — including Qwen3.5-2B (16.3 points). This result continues to uphold the Law of Intelligence Density put forward by MiniMax, which observes that the intelligence density of large language models roughly doubles every 3.5 months. The Base version of the model was pre-trained with ForgeTrain, an in-house AI training framework developed by MiniMax. ForgeTrain is the world’s first production-grade training framework fully written by AI. When quantized to INT4 precision, the model’s weight file is only 0.5 GB in size, allowing it to run on more than 90% of common end devices such as mobile phones and web browsers. It also comes with native support for mainstream inference frameworks including vLLM, SGLang and llama.cpp.
+
 
 ### Integrated Deployment
 - Out-of-the-box inference scripts with pre-configured hardware and software parameters