Baseten, the AI infrastructure company valued at $2.15 billion, has introduced Baseten Training, a platform designed to revolutionize model training for enterprises. This move aims to reduce reliance on closed-source AI providers like OpenAI and enhance customization of open-source models. The platform offers a solution to the operational challenges of managing GPU clusters and cloud capacity planning, empowering companies to fine-tune AI models independently.
Driven by customer demand, Baseten’s expansion beyond inference services focuses on providing companies with full control over their training code, data, and model weights. The platform’s technical capabilities include multi-node training support, sub-minute job scheduling, and integration with Baseten’s Multi-Cloud Management system, setting it apart from hyperscalers like AWS and Google Cloud.
Early adopters have reported significant cost savings and latency improvements, showcasing the platform’s effectiveness for custom models. Baseten’s strategy of owning both training and inference stages aims to optimize the AI lifecycle, offering performance enhancements and operational efficiencies for enterprises.
As the industry shifts towards fine-tuning open-source models, Baseten’s developer-centric approach and commitment to technical excellence position it as a key player in the evolving AI infrastructure landscape.
Source: VentureBeat