Auto Model: Stop Thinking About Which AI Model to Use
Intelligent model routing that picks the right AI model for you, automatically.
You’ve probably been there. You’re deep in a coding session, architecting a new feature with a powerful reasoning model. Then you switch to implementation and realize you’re still burning through a ton of tokens for straightforward code generation. So you switch to a cheaper model. Then back to the premium one for debugging a complex issue. Then back again for some coding - it’s exhausting and shouldn’t be necessary. The whole dance gets old fast.
That’s why we’re introducing Auto Model for Kilo.
The Model Selection Load
Different tasks need different models. Planning requires deep reasoning. Implementation needs speed and efficiency. But choosing between them adds friction every time you switch contexts, and getting it wrong means either wasting money or sacrificing quality…which is exactly what Auto Model eliminates.
How It Works
Select kilo/auto once and you’re done.
Auto Model uses Kilo Code’s existing work modes as signals to route each request to the optimal model:
Planning and reasoning tasks (Architect, Orchestrator, Ask, Plan modes) → Claude Opus 4.6 for complex system design and multi-step thinking
Implementation work (Code, Build, Debug, Explore modes) → Claude Sonnet 4.5 for fast, cost-effective code writing and editing
You just switch modes like you normally would. Auto Model handles the model selection in the background.
What This Looks Like in Practice
Say you’re building a new authentication system. You start in Architect mode, Auto Model routes to Claude Opus 4.6 for planning and architecture design. Switch to Code mode for implementation, Sonnet 4.5 takes over for fast execution. Hit a bug in Debug mode, Sonnet traces and fixes it. Need to rethink the architecture? Back to Opus 4.6.
Cost Breakdown
The cost efficiency is just as real as the convenience.
Claude Opus 4.6 is more expensive than a model like Sonnet 4.5. If you use premium options for everything, you’re paying top dollar for routine tasks. If you use only fast models, you’re leaving quality on the table when you need deeper reasoning.
Auto Model gives you both: frontier-level thinking where it matters, cost-effective execution where it doesn’t.
What’s Next
Today, Auto Model routes between Anthropic’s models: Opus 4.6 for reasoning, Sonnet 4.5 for implementation.
In the coming weeks, we’re expanding to route requests across more of Kilo model catalog, including free models and more cost effective ones.
The goal: automatically optimize your quality-to-cost ratio for every single task, invisibly and continuously.
Getting Started
Open Kilo Code in your IDE (VS Code, JetBrains, or CLI)
Select kilo/auto from the model dropdown
Work normally
From there, Auto Model handles model selection in the background.
(Note: Auto Model requires VS Code/JetBrains extension v5.2.3+ or CLI v1.0.15+. On older versions, kilo/auto defaults to a standard model for all requests.)
The Philosophy
We think that model selection should be invisible. As we’re seeing new models drop every week, we don’t want developers to have to become experts in every AI model. You should be able to focus on architecture or debugging and not thinking that hard about what model to use…while also having the freedom to take advantage of the latest models from anywhere.
Auto Model is our first step toward AI model selection that’s fully automatic, continuously optimized, and invisible. Start with kilo/auto today, and let the system handle the models while you handle the code.
Ready to try Auto Model? Read the docs or install Kilo Code and select kilo/auto to get started.