Stability AI and its CarperAI lab have launched Stable Beluga 1 and Stable Beluga 2, advanced open access Large Language Models (LLMs) with exceptional reasoning capabilities.
The models are based on the LLaMA framework, with Stable Beluga 1 using the 65B version and Stable Beluga 2 using the 70B version, both showing industry-leading performance.
Training used a synthetically-generated dataset, reducing cost and carbon footprint and resulting in high performance across various benchmarks.
Stable Beluga 2 leads the Open LLM Leaderboard as of July 27th, 2023, with Stable Beluga 1 also ranking highly at fourth place.
These models aim to advance research, improve natural language understanding, and enable complex tasks, with a commitment to an open and collaborative future in AI.

Stability AI is a startup focused on creating state-of-the-art artificial intelligence models and promoting open-source collaboration. It provides a platform for developers to contribute to and build applications using their foundational model, Stable Diffusion XL. They are pushing the boundaries in audio diffusion to generate high-quality music and sound effects, and they maintain a commitment to open access language models, competing with industry leaders. Stability AI has also developed the first open generative AI video model based on their image model. They emphasize community involvement and offer updates through a newsletter subscription.

Stability AI

Stability AI offers the Stable Diffusion model for images, audio, and language.

Stability AI Launches Stable Beluga Language Models

Recent Product News