A hands-on experiment fine-tuning Gemma 3 27B using synthetic data from Sonnet 4, achieving performance that rivals GPT-4o on specific tasks while being smaller and cheaper to run.

Most teams are curious about fine-tuning and how it might help their AI products, but they don't know what to expect from the process, how to measure success, or where to start.

Qwen 3, Gemma 3, Evals Guides, and Fine Tuning Updates!

Learn how to fine-tune LLMs without coding in under 20 minutes. Turn existing prompts into custom models that are faster, cheaper, and often better than prompting larger models.

Vertex, Azure, QwQ, Gemma, Hugging Face, W&B, Anthropic, Import CSV, and More!

Powerful Evaluation Toolkit & Support for Distilling Sonnet 3.7 Thinking

Easily create your own LLM evals in minutes. Use powerful evaluation methods like LLM-as-Judge and G-eval from Kiln's UI without coding.

A new release, faster synthetic data, bug fixes, and Kiln news.

That was fast. Thank you for the support!

Create your own reasoning model in under 30 minutes with Kiln AI.

The Vector Institute, a research lab founded by Geoff Hinton, has started using Kiln for their AI research.