Optimal Model Selection for Cost and Performance

Optimal Model Selection for Cost and Performance | Generated by AI

Home 2025.09

Looking at your Claude Code configuration, you want to optimize for both cost and performance. Here are some recommendations for choosing “cheap and powerful” models in your router setup:

Best Cost/Performance Models from your list:

For most tasks (default):

deepseek/deepseek-chat-v3.1 - Excellent performance at very low cost
deepseek/deepseek-r1 - Great reasoning capabilities, still cost-effective

For background tasks:

deepseek/deepseek-chat-v3.1 - Already a good choice
google/gemini-2.5-flash - Fast and inexpensive

For thinking/reasoning:

deepseek/deepseek-r1 - Specifically designed for reasoning tasks
z-ai/glm-4.5 - Your current choice is decent

Updated Router suggestion:

"Router": {
  "default": "openrouter,deepseek/deepseek-chat-v3.1",
  "background": "openrouter,google/gemini-2.5-flash", 
  "think": "openrouter,deepseek/deepseek-r1",
  "longContext": "openrouter,moonshotai/kimi-k2",
  "webSearch": "openrouter,google/gemini-2.5-flash"
}

Key considerations:

DeepSeek models offer the best price/performance ratio currently
Gemini 2.5 Flash is very fast and cheap for simpler tasks
Keep Kimi for long context since it handles 200k+ tokens well
Gemini 2.5 Pro if you need more capability than Flash

You can check current pricing on OpenRouter’s website to compare exact costs, but DeepSeek models are typically 10-20x cheaper than GPT-4 level models while offering competitive performance.

Back Donate