Model Colour + Label System
KForge uses a simple colour-based labelling system to help users quickly understand
the expected cost, risk, and usage profile of a model.
These labels are designed for human decision-making, not billing precision.
Label Overview
| Label |
Cost Meaning |
Typical Use |
| π΅ Free β Sandbox |
Free models |
Testing / experimentation |
| π’ Paid β Sandbox |
Low-cost paid models |
Testing / experimentation |
| π‘ Paid β Main |
Mid-cost paid models |
Default development work |
| π΄ Paid β Heavy |
High-cost paid models |
Complex / critical tasks |
| βͺ Unknown |
Pricing unclear |
Use with caution |
π΅ Free β Sandbox
Meaning
- No billing required
- Provider-controlled limits
- Often rate-limited or capability-capped
Usage
- Safe to experiment
- Ideal for quick tests and learning
- βMess around modeβ
π’ Paid β Sandbox
Meaning
- Paid, but cheap
- Low per-request cost
- Designed to be disposable
Usage
- Safe for testing and iteration
- OK to spam
- βMess around mode (paid)β
Examples:
Tooltip:
Low-cost paid model. Suitable for testing and experimentation.
π‘ Paid β Main
Meaning
- Paid
- Reasonably priced
- Balanced capability vs cost
Usage
- Default development work
- Your day-to-day workhorse
- Be mindful, not paranoid
Examples:
gpt-4.1-mini
- strong gateway / general-purpose models
Tooltip:
Balanced paid model for daily development.
π΄ Paid β Heavy
Meaning
- Expensive
- High capability
- Optimised for quality, not volume
Usage
- Complex reasoning
- Accuracy-critical tasks
- Think before spamming
Examples:
- high-end models
- advanced reasoning models
Tooltip:
High-cost model. Use when accuracy matters.
βͺ Unknown
Meaning
- Pricing unclear
- Custom endpoints
- Gateway- or provider-dependent
Usage
- Use with caution
- Assume paid until proven otherwise
Important Notes
- These labels are informational only
- Billing, limits, and availability are determined by the provider
- Users assign labels manually per model
- Labels describe expected usage patterns, not guarantees
KForge does not enforce cost or block usage.
The goal is clarity, not restriction.