Kiln Desktop - v0.19.1

What's Changed

• Model Library: Browse all available models and their supported capabilities in our new model library.
• New AI Providers: Cerebras, SiliconFlow.cn
• New Models: GPT-5, GPT OSS, GLM, Qwen 3 2507, MiniMax, Hunyuan, Pangu Pro
• Better internationalization: Allow non-ascii characters in our model names, names and descriptions.
• And More: fine-tuning UI improvements, native file-picker for importing projects, design updates, bug fixes

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/16840409052
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/16840408959

Full Changelog: v0.18.1...v0.19.1

Kiln Desktop - v0.18.1

Overview: Evals V2, Synthetic Data Gen V2, Kiln 'Issues' New Models, and More!

• Evals V2: Our evals are now more powerful and easier to use. Features include simplified eval creation, improved prompts, eval templates, a new "compare" view to find the best model, and more!
• Synthetic Data V2: The data generation tool has been redesigned to cover more use cases, including fine-tuning, eval data, issue resolution, and more. The UI has been improved and simplified.
• Introducing 'Issues': Kiln Issues are like software bugs, but for AI systems. Describe an issue, create an eval for it, generate data to reproduce it, and fix it with tuning or prompting.
• New Models: Added Kimi K2, Grok 4, Llama 4, R1 0528, and more! We can now publish new models over the air, so no more waiting for the next release.
• Improved UI: Added type-to-filter functionality in dropdowns like model selection, plus more model suggestions.
• Bug Fixes: Fixed an issue related to how model outputs are sent to eval judges.

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/16310037457
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/16310037455

Full Changelog: v0.17.0...v0.18.1

Kiln Desktop - v0.17.0

• Synthetic Data UX Improvements: Your synthetic data session is now preserved; you can leave and come back without losing data. Visual updates make synthetic data easier to use.
• Powerful new logs: View the exact requests/responses sent to/from LLM providers. You can open the logs folder from settings in app. Useful for understanding workflows, debugging, or productionizing Kiln tasks.
• New parameters: top_p, temperature, structured output mode: specify key LLM parameters both on the run screen, and during evals when trying to find the best method of running your task. Structured output mode let's you try different methods of enforcing data structure (JSON schema, tool calls, instructions)
• Much more: bug fixes, usability improvements, model list updates, CoT in fewer messages.

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/15665812199
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/15665812198

Full Changelog: v0.16.0...v0.17.0

Kiln Desktop - v0.16.0

• New Models: Qwen 3 and Gemma 3 including the ability to run /no_think or /think for Qwen 3
• Major Redesign for Evals: Creating an eval now walks you through each step of the process with an intuitive UI. Building your dataset, rate golden data, compare eval algorithms, and more. You can create state of the art evals without prior experience.
• Improved fine-tuning: it's now easier to create fine-tuning datasets and manage existing datasets.
• Many quality of life improvements: new prompt generator, suggested models for evals/data-gen, improved UI controls, simpler setup

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/15085724878
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/15085835381

Full Changelog: v0.15.0...v0.16.0

Kiln Desktop - v0.15.0

⬇️ Download links in the "Assets" section below ⬇️

This release includes:

• New Models: GPT 4.1 Full/Mini/Nano, o4 mini, o3, and Gemini 2.5 Pro/Flash.
• Gemini Fine-Tuning: connect a Google Vertex account to fine-tune Gemini models.
• Faster Synthetic Data Gen: generate data for all topics in a single click.
• New UI which allows manually repairing data
• Fix a bug breaking some chain-of-thought prompts

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/14617773103
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/14617773044

Full Changelog: v0.14.0...v0.15.0

Kiln Desktop - v0.14.0

⬇️ Download links in the "Assets" section below ⬇️

This release adds a ton of new integrations including 5 new AI providers, 5 new inference models, 3 new fine-tune base-models

• Added over 60 new fine-tuneable models like QwQ, Deepseek V3, Llama 3.3, Qwen 2.5, and more. These use Fireworks AI, and don't require setting up custom GPUs.
• Added the ability to deploy on-demand inference servers for Fireworks AI. These deployments will automatically scale to zero servers after 5 minutes of inactivity, making it easy to keep costs down. When you go to use a model, a server will automatically start for you.
• Bug fix blocking connecting Anthropic keys to Kiln

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/14285733084
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/14285733097

Full Changelog: v0.13.2...v0.14.0

Kiln Desktop - v0.13.2

⬇️ Download links in the "Assets" section below ⬇️

Note: v0.13.2 is a minor update from v0.13.1 with a bug fix for Ollama. The remaining release notes are from v0.13.1

This release adds a ton of new integrations including 5 new AI providers, 5 new inference models, 3 new fine-tune base-models

• New AI Providers: Gemini API, Vertex AI, Hugging Face, Azure OpenAI and Together.ai
• New Inference Models: QwQ, Gemma 3, o1, o3, Phi 4 Mini
• New fine-tuning host: Together.ai which adds new serverless fine tuning models: Qwen 2.5 14B, Qwen 2.5 72B, Llama 3.2 1B
• Features:
  • Import existing datasets via CSV (thanks leonardmq)
  • Weights and Biases integration for fine tuning metrics
  • Allow editing titles/descriptions throughout the UI
  • Add UI for deleting evals, tasks, custom prompts, and other data model items directly from the UI
• Core tech
  • Migrate to LiteLLM, removing the last of LangChain
  • Python 3.13 support (and ready for 3.14)

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/14007063283
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/14007063288

Full Changelog: v0.12.1...v0.13.2

Kiln Desktop - v0.13.1

⬇️ Download links in the "Assets" section below ⬇️

This release adds a ton of new integrations including 5 new AI providers, 5 new inference models, 3 new fine-tune base-models

• New AI Providers: Gemini API, Vertex AI, Hugging Face, Azure OpenAI and Together.ai
• New Inference Models: QwQ, Gemma 3, o1, o3, Phi 4 Mini
• New fine-tuning host: Together.ai which adds new serverless fine tuning models: Qwen 2.5 14B, Qwen 2.5 72B, Llama 3.2 1B
• Features:
  • Import existing datasets via CSV (thanks @leonardmq)
  • Weights and Biases integration for fine tuning metrics
  • Allow editing titles/descriptions throughout the UI
  • Add UI for deleting evals, tasks, custom prompts, and other data model items directly from the UI
• Core tech
  • Migrate to LiteLLM, removing the last of LangChain
  • Python 3.13 support (and ready for 3.14)

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/13956877583
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/13957266813

Full Changelog: v0.12.1...v0.13.0

Kiln Desktop - v0.12.1

⬇️ Download links in the "Assets" section below ⬇️

New in this release Evals - a comprehensive evaluation toolkit:

• Build powerful SOTA evals (G-Eval, LLM as Judge)
• Check eval correlation to human preferences (Kiln Ratings) to find the best evaluator
• Synthetically generate eval datasets with Kiln Synthetic Data Gen
• Use the analysis tools to find the optimal prompt+model for your task
• Automatic eval: Kiln will automatically build an eval for any Kiln task using your task definition
• Templates for common eval use cases: bias, toxicity, jailbreaking, maliciousness, factual correctness.

Other new features include:

• Support for distilling (fine-tuning) an open model from Sonnet 3.7 thinking
• New Built-In Models: Sonnet 3.7, Dolphin 2.9 8x22B, and Grok
• Improved logging (thanks to @leonardmq)
• ARM Linux builds now included

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/13638728833
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/13638728765

Full Changelog: v0.11.1...v0.12.1

Kiln Desktop - v0.11.1

⬇️ Download links in the "Assets" section below ⬇️

This release focuses on making Kiln easier to use, across lots of small fixes.

New features include:

• Synthetic data generation is 25x faster
• Windows Code Signing which means fewer "Unrecognized app" prompts, and virus false positives
• Resolve a bugs causing a blank screen on some Windows 11 machines
• Improved Ollama support when using structured data (JSON)
• You can disconnect AI providers from settings UI, without editing config files
• Improved docs, and links to the docs from within the app
• Add the ability to clone a task, copying all it's fields
• Show synthetic dataset topic in the datasets view

CI Build Source for this release Mac and Linux: Kiln-AI/Kiln/actions/runs/13234579948 and Windows: Kiln-AI/Kiln/actions/runs/13234579954

Full Changelog: v0.10.1...v0.11.1