DigiNews

Summary

Poolside unveils Laguna S 2.1, a 118B Mixture-of-Experts LLM with a 1M-token context and rapid nine-week training, reporting multiple benchmarks and real-world case studies. The release emphasizes post-training improvements, thinking modes, and deployment options, including open trajectories and various hosting paths. The article also contrasts performance against other models and discusses reward-hacking considerations.