Blog
Guides and pricing notes.
How to call frontier models through one key, and what it costs. Subscribe with RSS.
Chat Completions vs Messages vs Responses: which LLM API format?
The three request formats behind most LLM APIs, side by side: required fields, where the system prompt goes, how to read the reply, and when to pick each.
Claude Fable vs Opus vs Sonnet vs Haiku: API cost per task
What one task costs on each Claude tier, for three task shapes, with October 2026 list prices. Plus a simple rule for picking the tier.
Handling 429 and 503 errors from an LLM API
Which LLM API errors to retry, how long to wait, and how to fall back to another model when there is no capacity. With Python code for the OpenAI SDK.
LLM API pricing explained: what per-million-token prices cost
How input and output token prices add up, a worked monthly estimate for Claude, GPT and Gemini, and five ways to cut the bill.
How to call Claude, GPT and Gemini with one API key
Point the OpenAI or Anthropic SDK at one base URL, pick a model id, and call Claude, GPT and Gemini models with the same key. Python, TypeScript and cURL.