You’re overpaying your AI models.
Auto-routing picks the cheapest one that still works.
Kilo just shipped **Auto Efficient** — a new tier in their Auto Model lineup.
Instead of locking you into a single model (or asking you to switch manually), a lightweight classifier reads your session in context and routes each request to the cheapest model proven accurate enough for the work.
The router runs continuously. Ask for a quick rename → it picks something fast and cheap. Plan a tricky migration → it reaches for a stronger model. Same setting, different work, different price.
Three things I like about how it’s built:
• It routes on **KiloBench** — a public coding benchmark Kilo runs continuously — not vendor marketing copy.
• It’s session-aware. Once it picks a model that works for your thread, it stays there. No mid-conversation whiplash.
• When it’s not sure, it falls back to **Balanced** (a capable paid model). So the quality floor is hard, not “best effort.”
There’s a dial too. One setting squeezes the most value from every dollar. The other leans toward the strongest proven model in the pool. You set it per project based on whether you care more about the bill or the bar.
And it’s still optional. Step out anytime — pick a specific model by hand, bring your own provider key, or run a local model through Ollama or LM Studio. The router is there when you want it and gone when you don’t.
The bigger takeaway: most teams are paying frontier prices for routine work. Smart routing shifts that spend to where it actually moves the needle.
Available now in Kilo — needs VS Code or JetBrains extension v5.2.3+, or CLI v1.0.15+.
#AI #CostOptimization #LLMRouting #Claude #Automation #DevTools #AICoding #AIAgents #IndieHacker #SoftwareEngineering
Video Source
About the Author
