What's New in V1.1.1
This release includes Kiln Evals V2: a major improvement to our evaluation system in Kiln. Plus Code Tools, Featherless AI, new models, and more.
- Evals V2
- More Eval Types: Code evals, tool calling evals, regex evals, turn counting, and more. All faster and cheaper than LLM as judge. Or mix and match: a code judge that uses LLMs when needed.
- Better eval efficiency: reuse traces across many evals.
- Code Tools: create tools (like MCP tools) using Python without leaving Kiln.
- Featherless AI: a new AI provider with thousands of models to choose from.
- New Models: Opus 5, Qwen 3.8, Kimi K3 (more providers), New DeepSeek V4 Flash and Pro, Grok 4.5, Qwen 3.7 Max, Mistral Medium 3.5, Thinking Machines Inkling, Gemini Flash Lite/3.6/3.7, Muse Spark, MiniMax M2.7
- And More: Higher tool call limit, bug fixes, UI updates, and more.
CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/32419561133
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/32419561068
Full Changelog: v1.0.4...v1.1.1