AI Briefing#sakana-ai
The latest AI and developer news about #sakana-ai, with the original source and a short summary.
The latest AI and developer news about #sakana-ai, with the original source and a short summary.
MASS based on Qwen3.6-27B improves research benchmark performance by 1.2–1.6x, while SW benchmarks decline
The MASS framework enables a single LLM to act as optimizer, evaluator, and agent, achieving a 69.9% win rate over the base model in Round 2 on test tasks.
Sakana AI proposed PC-ALM, which resolves the signal attenuation issue in existing Predictive Coding (PC), successfully training 1000-layer neural networks.
Sakana AI has released Fugu Ultra v2, which employs a multi-agent routing approach rather than a single model.
Sakana AI's DiffusionBlocks technique, presented at ICLR 2026, significantly reduces memory usage during deep learning training.
Sakana AI has released Fugu-Ultra v1.1, a multi-agent orchestration model with significantly improved performance by integrating the latest frontier models.
Sakana AI has published an Error Diffusion technique that is trainable while adhering to Dale's principle without backpropagation.
Sakana AI has unveiled a distributed hardware system that recognizes its own shape and detects damage without any central control.
With U.S. export controls limiting access to Anthropic's models, Asian companies are launching alternative models to capture the market.
Amid U.S. export restrictions, Japan's Sakana AI and China's 360 have responded by launching high-performance AI models.
Sakana AI has launched the Fugu API, which dynamically orchestrates closed and open-source models for complex tasks.
Sakana AI's multi-model collaborative model, Fugu Ultra, has been released via Vercel's AI Gateway.