WORLD3's RouterLink integrates 21 new AI models, further expanding its decentralized model routing service.

According to Mars Finance, RouterLink, WORLD3's decentralized large-scale model routing service platform, announced the completion of a large-scale model expansion, integrating 21 new AI models from 10 providers at once. This update covers four core categories: flagship inference models (such as GLM-4.7, Gemini 3 Pro, Nova 2.0 Pro, Kimi K2 Thinking, MiMo-V2-Pro), high-speed response models (such as Gemini 3 Flash, Kimi K2 Flash, GLM-4.7-Flash, Nova 2.0 Lite, Seed 1.6 Flash, MiMo-V2-Flash, MiniMax M2.1 Flash), multimodal generative models (such as GPT Image 1.5, Seedream 4.5, Maverick 17B (Llama 4), Seed 1.6), and programming-specific models (such as Qwen3 Coder 480B, Qwen3 Coder 32B, MiniMax M2.1, Llama 3.1 8B / 70B Instruct), aiming to support flexible routing and invocation of AI agents in complex application scenarios. The WORLD3 team stated that developers can access these cutting-edge models with zero migration cost through RouterLink's unified API. The platform has also optimized the UI experience for model selection and discovery to improve development efficiency. Currently, all newly added models are officially online.