Together AI now supports fine-tuning of OpenAI's newly open-weight gpt-oss-20B and gpt-oss-120B models (Apache 2.0 licensed). The platform simplifies distributed training into a three-step process — upload dataset, configure parameters, launch job — handling LoRA training, data validation, and deployment automatically. Fine-tuned models can be deployed to dedicated endpoints with 99.9% uptime SLA and SOC 2 compliance. Supported configurations include 16K context for SFT and 8K for DPO. Key benefits highlighted include model stability (no vendor-driven drift), cost efficiency of smaller specialized models, and freedom from managing GPU clusters.
