unslothai/unsloth Adds gpt-oss and Kimi Local Fine-Tuning — First Training Framework to Support OpenAI Open-Weight Models

GitHub·medium signal

Unsloth, the 56K-star fine-tuning optimization library, updated its description to explicitly include gpt-oss and Kimi-K2.5 in its locally-runnable model list alongside Qwen, DeepSeek, and Gemma. This makes Unsloth the first major training framework to enable fine-tuning of OpenAI's newly open-sourced gpt-oss-120B locally on consumer hardware. For builders, this means custom fine-tunes on frontier-class open weights are now practical within days of the model release.