Choosing the right AI model can save you 9x on API costs. Here is the data.
| Model | Input | Output | vs GPT-4o |
|---|---|---|---|
| GPT-4o | $2.50 | $10.00 | baseline |
| Claude 3.5 Sonnet | $3.00 | $15.00 | 1.5x more expensive |
| DeepSeek V3.1 | $0.27 | $1.10 | 9x cheaper |
| DeepSeek R1 | $0.55 | $2.19 | 5x cheaper |
| Llama 3.1 8B (local) | $0 | $0 | Free |
Do not pick one model — route to the best model per request: chat to local, code to DeepSeek, reasoning to DeepSeek V3.1, vision to GPT-4o, long context to Claude.