Opus…Sonnet…Haiku?

Knowing these are different models one can use, vs, understanding when to use these models can be a complete game changer.

Opus 4.6 - Used when getting it right matters more than speed/cost. This is the O’g, it will build you a world just with one sentence. Very capable, very powerful. Useful for really complex, challenging tasks. Has a large context window. Compared to previous model, this version is better at generating UI’s, designs etc.

Sonnet 4.6 - It too can build you a world, just maybe not as big. Still very beautiful. Use it to handle everyday, slightly challenging tasks - tasks that do not require intense thinking, deepdives etc. Dont get me wrong, its still very capable, however it will be better to use Opus for tasks that need the extra brain power. Good balance between performance vs cost vs speed.

Haiku 4.5 - This little guy packs a punch. Think of it as the little brother that is always there to help. Ever ready to get things done just might not always be the best outcome. There are tasks it excels at. Its optimized for speed and cost. Use this model for mundane, repetitive tasks. Useful for quick, cheap questions/answers. e.g. customer service agent, chatbot.

Useful for sub-agents. Break down complex tasks into smaller, byte-sized bits and hand it over to Haiku powered sub-agents to work on. This reduces costs, time. Can be used as a routing agent, incoming AI request, it can analyse it and determine which model is best to handle it.

Claude Code vs, Claude API vs, Claude AI

Claude AI: This is the UI, that is available for everyone to use. Friendly frontend, with everything at your fingertips - makes use of a subscription. (free tier available).

Claude Code: Available via the CLI(command line interface) as a CLI tool. It brings Claude directly into your terminal, so you can build, edit, and debug code without ever leaving your development environment.

Claude Console: Lets you manage your access to API. Test prompts , monitor usage, set spending limits.

Claude API: Allows developers to integrate claude AI into whatever they are building. Gives them access to claude models via an API. Uses pay-as-you-use model instead of a subscription. This is separate from your Claude AI subscription/plan. It’s the foundation that everything else is built on — Claude.ai itself runs on it, and Claude Code uses it under the hood too

Claude SDK: Allows developers to embed claude into whatever they are building. Built around the API - the API is the connection, the SDK is the toolbox that makes using that connection easier

So Who’s it for?

Claude AI → end users, professionals, everyday people.

Claude Code → developers who want AI embedded directly in their dev workflow.

Claude Console/API/SDK → developers/companies building their own products with Claude inside. Managed via Anthropic console.