Together AI launches post-training service for custom AI models

Together AI has launched Together Post-Training, a new service designed to help developers customize open-weight AI models using reinforcement learning (RL) and supervised fine-tuning (SFT). The service provides reserved GPU capacity, allowing users to manage training data, objectives, and experiment steps while the platform handles the underlying infrastructure.

Key features include support for multiple LoRA training experiments on the same reserved GPU capacity, full-weight training for specific models, and asynchronous reinforcement learning. The service also allows customers to serve several trained LoRA adapters through a single dedicated inference deployment. The initial model lineup includes Muse Spark Open, GLM 5.3, Kimi K3, and Qwen 3. 8-27B.

Together AI is offering an introductory rate of $9 per B200 GPU-hour for the first eight weeks. The launch follows the company’s $800 million Series C financing in July and commitments for over 500 megawatts of computing capacity. The service aims to enable organizations to train specialist agents that match frontier model quality at a lower cost and runtime compared to general-purpose models.