How to reduce LLM API costs

Ten practical levers

• Route simple requests to a small model.
• Cap reply length with max_tokens.
• Shorten and reuse system prompts; cache them where supported.
• Trim old conversation turns.
• Use low thinking levels for routine tasks.
• Batch classification into fewer, larger prompts.
• Store results you will reuse instead of regenerating them.
• Embed documents once and store the vectors.
• Pick image and video resolutions you actually need.
• Watch the usage field in every response.

Example

from openai import OpenAI

client = OpenAI(api_key="YOUR_TOKENREX_KEY", base_url="https://tokenrex.ai/api/public")

resp = client.chat.completions.create(
    model="gpt-6-luna",
    messages=[{"role": "user", "content": "Summarise in 2 lines"}],
    max_tokens=120,
)
print(resp.choices[0].message.content)

Questions

Do I need a separate key for each provider?

No. One TokenRex API key works for every enabled model; change only the model ID.