arXiv:2605.27605v1 Announce Type: new Abstract: We present Laguna M.1 and Laguna XS.2, two Mixture-of-Experts foundation models built for long-horizon, agentic coding: M.1 has $225.8$B total parameters ($23.4$B activated per token) and XS.2 has $33.4$B total ($3$B activated). Both models were trained from scratch end-to-end inside the same internal system that we refer to as our Model Factory: a tightly-integrated stack of versioned data, training, evaluation, and inference components that turn model development into an industrial process. We describe the principles and design choices of the M
The continuous evolution of AI foundation models, particularly in agentic capabilities and long-horizon tasks, necessitates frequent updates on new architectures and training methodologies, like those presented in this technical report.
This report details new Mixture-of-Experts (MoE) models specifically designed for agentic coding, indicating a significant step towards more autonomous AI agents that can handle complex, multi-step programming tasks.
The development of MoE models trained within an 'industrial process Model Factory' suggests that AI model development is becoming more standardized and scalable, potentially accelerating the deployment of highly capable AI agents.
- · AI software developers
- · Companies adopting AI agents for coding
- · Cloud providers offering agentic AI services
- · Traditional software development firms slow to adapt
- · Individual coders whose tasks are automated
More efficient and autonomous software development cycles due to advanced AI coding agents.
Increased demand for specialized hardware and infrastructure to support the training and deployment of large MoE models.
A potential restructuring of the software engineering workforce, shifting from manual coding to AI agent supervision and architecting.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI