Runpod AI field notes: December 2025

In early December, Mistral AI and Nvidia made significant strides in the open model ecosystem, with Mistral AI releasing two open models: Mistral Large 3, a mixture-of-experts model featuring 41 billion active parameters, and Devstral 2, optimized for coding and tool use. Both models are available under the Apache 2.0 license, allowing for customization and deployment on personal GPU infrastructures. Concurrently, Nvidia introduced the Nemotron 3 family of open-source models, which are optimized for Nvidia GPUs and excel in various tasks, offering faster performance than some competitors. Nvidia also announced the acquisition of SchedMD, the company behind the widely-used Slurm scheduler, reinforcing its commitment to open-source infrastructure for AI and HPC clusters. These developments indicate a growing trend towards open models and infrastructure, enabling developers to experiment and build AI systems with greater flexibility and transparency, while reducing dependency on closed platforms.