Sakana AI, led by former Google researchers David Ha and Llion Jones, is reshaping the landscape of AI development through its innovative technique known as ‘model merging.’ By merging multiple AI models, Sakana AI has achieved a significant breakthrough in creating more powerful and adaptable systems.
These new models exhibit remarkable advancements, including improved performance metrics, enhanced adaptability across diverse tasks, and unmatched robustness in various domains, according to reports. Through extensive experimentation, they’ve successfully evolved Japanese Large Language Models (LLMs) and Vision-Language Models (VLMs) to outperform existing benchmarks. Additionally, their method extends to Image Generation Diffusion Models, resulting in high-quality Japanese-capable SDXL models. The introduction of three Japanese language models, EvoLLM-JP, EvoVLM-JP, and EvoSDXL-JP, with two of them being open-sourced, represents a significant leap in AI development. Sakana AI aims to establish Tokyo as a prominent centre for AI innovation, following the footsteps of OpenAI in San Francisco and DeepMind in London.