📌 Strategy: Uses credits first across all models, then falls back to tokens.
📌 Best for: Users who prefer to spend their credits first before consuming earned tokens.
How it Works #
- If the primary model (e.g., DeepSeek-7B) lacks credits, the system checks the next model in the priority list.
- If no models have credits available, the system falls back to tokens, starting from the highest-priority model.
Example Scenario #
Model | Priority | Available Tokens | Credit Needed | User Credits |
---|---|---|---|---|
DeepSeek-7B | 1 | 100 | 20 | 50 |
LLaMA 3-8B | 2 | 0 | 15 | 10 |
Mistral-7B | 3 | 50 | 10 | 5 |
Final Selection: #
✅ DeepSeek-7B → Uses Credits
✅ LLaMA 3-8B → Uses Tokens
✅ Mistral-7B → Uses Tokens
📌 Why? Unlike previous strategies, credits are always used first, ensuring that token reserves are only used when necessary.