Researchers have introduced Laguna M.1 and Laguna XS.2, two Mixture-of-Experts foundation models designed for long-horizon, agentic coding tasks. Laguna M.1 features 225.8 billion total parameters with 23.4 billion activated per token, while Laguna XS.2 has 33.4 billion total parameters with 3 billion activated. Both models were developed using an internal "Model Factory" system, which integrates data, training, evaluation, and inference components for an industrial approach to model development. The models demonstrate competitive performance against state-of-the-art open models on agentic software engineering and terminal benchmarks, with Laguna XS.2 weights now available under the Apache 2.0 license. AI
IMPACT Introduces new open-source models for agentic coding, potentially advancing development in automated software engineering.
RANK_REASON The cluster describes the release of new AI models detailed in a technical report, including their architecture, training process, and benchmark performance. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
