Efficient
NewSession-aware routing for smarter spend.
Routes using live session classification.
See benchmark vs Opus 4.8Automatic model routing
The right model for every task
Auto Model routes each request to the model that best fits the work, so you spend less. Optimize your workflow with smart model orchestration.
AnthropicOpenAIGooglexAIDeepSeekQwenMoonshotMistralMiniMaxTencent HunyuanZ.aiInclusionAIKwaipilotPoolsideStepFunNVIDIA
Model freedom without model management
Kilo gives you access to 500+ hosted, BYOK, and local models. Auto Model turns that choice into smart orchestration: optimize cost and workflow by letting Kilo route each session to the right model while you stay focused.
Auto Model routing pipeline
1. Set Tier
Efficient
Frontier
Balanced
Free
2. Start Session
Start Session
Classifies session intent
3. Kilo Routes
Best Model for Task
Routed via benchmarks & usage data
500+ hosted models, BYOK, and local models are still available — Auto Model is the default, not a lock-in.
New: Kilo Auto Efficient
Auto Efficient delivers near-frontier coding performance at dramatically lower cost.
Auto Efficient delivers 71% of published frontier completion at 72% lower cost on KiloBench. See how it compares with Claude Opus 4.8, Claude Sonnet 4.6, and GPT-5.5 on completion rates, cost per attempt, and one-shot app demos.
Lower cost
vs frontier average
KiloBench
completion benchmark
Read the efficient vs frontier analysis
How Auto Model works
Auto Model combines session-aware classification with benchmark-proven model selection, so every request lands on the right model for the work.
01
Applies your tier
Efficient, Frontier, Balanced, or Free defines the capability and cost envelope. Each tier uses a different routing strategy.
02
Classifies the session
A classifier model receives your session data and assigns it a session type from a deterministic list.
03
Selects benchmark-proven models
Based on the type of work classified and benchmarks Kilo runs, Auto Model picks the best model for your session.
04
Preserves control
Use Auto Model by default, or manually select hosted, BYOK, or local models when you prefer.
Choose your routing tier
Four tiers cover every tradeoff — from session-aware benchmark routing to maximum capability, cost-effective paid routing, or no-cost free-model routing.
Frontier
Maximum capability when the work needs it.
Routes to the latest and most capable paid models.
Balanced
Capable paid routing without frontier prices.
Routes all modes to a cost-effective paid model.
Free
Free and experimental models providers can offer at no cost.
Routes to the best free models available.
Related
Gateway gives Auto Model reach
Kilo Gateway supplies the 500+ hosted-model catalog and routing infrastructure behind Auto Model tiers.
Learn moreKilo Pass helps credits go further
Kilo Pass gives you extra free inference every month, so Auto Model has more room to route work across the models that fit each session.
Learn moreLeaderboard shows real-world usage
See which models developers actually use across coding, planning, debugging, and agent workflows.
Learn moreKiloBench measures coding performance
Browse officially promoted benchmark results for AI coding models on Terminal Bench 2.0.
Learn moreModel control
You still control the model stack.
Auto Model does not remove choice. It gives you a smarter default while keeping manual model selection, BYOK, local models, and tier selection available whenever you want them.
- Hosted models
- 500+
- Local support
- Ollama, LM Studio
- BYOK
- Any provider key
How Auto Model fits your workflow
Set your tier
Choose Efficient, Frontier, Balanced, or Free based on your cost and capability goals.
Let Kilo route
Auto Model classifies each session and routes to the optimal model for the work — no manual switching required.
Override any time
Pick a specific model, use BYOK keys, or run local models with Ollama or LM Studio whenever you prefer.
Pair with Kilo Pass
Kilo Pass adds free inference every month, giving Auto Model more room to route across the right models.
Frequently Asked Questions
Everything you need to know about Auto Model
What is Auto Model?
Auto Model is Kilo's automatic model routing system. You choose a tier, and Kilo routes each request to a model strategy that fits the work instead of making you switch models manually.
Does Auto Model save money?
Auto Model helps control cost through smart model orchestration. It routes each session toward models that fit the work, so routine tasks do not need frontier-level spend and harder work can still use stronger models when your tier supports it.
What is the difference between Efficient, Frontier, Balanced, and Free?
Efficient uses live session classification for session-aware routing. Frontier routes to the latest and most capable paid models. Balanced routes all modes to a cost-effective paid model. Free routes to the best free models available, including free and experimental models providers can offer at no cost.
Can I still choose a model manually?
Yes. Auto Model is a default routing option, not a lock-in. You can still pick a specific hosted model, use BYOK provider keys, or run local models with Ollama or LM Studio.
Does Auto Model work with Kilo Gateway?
Yes. Kilo Gateway provides the 500+ hosted-model catalog and routing infrastructure behind Auto Model tiers.
Does Auto Model work with Kilo Pass?
Yes. Kilo Pass gives you extra free inference every month, so Auto Model has more room to route work across the models that fit each session.
How do the Kilo Leaderboard and KiloBench relate to Auto Model?
The Kilo Leaderboard shows real-world model usage across developer workflows, while KiloBench measures coding performance on benchmarked tasks. Together, they give visibility into the model signals behind smarter routing decisions.
Let Kilo choose the model
Start with Kilo Code and use Auto Model when you want model freedom without the manual switching overhead.