Skip to content
Oraicle Docs

Model Selection Strategy

Which model you pick changes both the answer and the bill. The full list, with live prices, is on oraicle.me/models.

Recommendations:

  • Start with smaller models - Begin development with models like openai/gpt-oss-20b or zai-org/GLM-4.7-Flash, which answer fastest and cost least per token.
  • Test before committing - Evaluate 2-3 models on your specific tasks before selecting one for production.
  • Match model to task - Use specialized models for specialized tasks:
    • Reasoning: deepseek-ai/DeepSeek-R1-0528 or moonshotai/Kimi-K2-Thinking
    • Multilingual content: Qwen/Qwen3-Next-80B-A3B-Instruct or zai-org/GLM-4.6
    • Long documents: deepseek-ai/DeepSeek-V4-Pro (1M context)
  • Consider context length needs - Only pay for extended context when your application requires it.