Together AI Expands Fine-Tuning: Tool Calling, Reasoning, Vision, and 100B+ Model Training Now Available

Together AI Blog·medium signal

Together AI shipped a major fine-tuning platform update adding native support for tool-call fine-tuning, reasoning model training, and vision-language model adaptation, plus the ability to fine-tune models at 100B+ parameter scale with up to 6x higher throughput versus previous infrastructure. The update includes job cost and ETA estimates, making large-scale fine-tuning operationally predictable for enterprise teams. For builders running agentic pipelines, tool-call fine-tuning in particular closes a key gap — the ability to train models to reliably invoke specific APIs without prompt engineering workarounds.