Sakana AI

SoftMatcha 2

SoftMatcha 2 is an ultra-fast and flexible search algorithm designed to perform searches over trillion-scale natural language corpora in under 0.3 seconds. It effectively handles semantic variations such as substitution, insertion, and deletion, making it a powerful tool for large-scale text analysis.

RePo-OLMo2-1B-stage2-L5

RePo-OLMo2-1B-stage2-L5 is an autoregressive language model developed by Sakana AI, based on the allenai/OLMo-2-0425-1B model. It has been continually pre-trained on 50 billion tokens from the allenai/dolmino-mix-1124 dataset using the RePo (Context Re-Positioning) technique, which aims to enhance the model's understanding of context by adaptively restructuring input sequences.

PC-ALM

PC-ALM (Augmented Lagrangian Predictive Coding) is a local-learning algorithm that serves as an alternative to backpropagation by training deep networks through local energy-minimization dynamics. It aligns weight updates with backpropagation by accumulating per-layer constraint errors into layer-local Lagrange multipliers, allowing for efficient training in deep, narrow network architectures.

Sakana Fugu

Sakana Fugu is a multi-agent orchestration system that dynamically coordinates a diverse pool of powerful models to tackle complex, multi-step tasks. It offers superior performance by assembling and orchestrating agents through efficient collaboration patterns, without relying on a single-vendor dependency. Fugu is designed to integrate seamlessly into workflows via a single API, providing flexibility in agent selection to meet data, privacy, compliance, or organizational requirements.

Fugu-Cyber

Fugu-Cyber is a specialized multi-agent orchestration model designed for complex cybersecurity tasks, including vulnerability verification and threat intelligence analysis. By functioning as a unified API endpoint, it dynamically manages specialized sub-agents to provide reliable, state-of-the-art security analysis without single-vendor dependency.