Claude Opus 5 Brings a Million-Token Context to the CLI

James Turner

You know what is wild, Lachlan? This episode of our show is actually sponsored by Jellypod AI, which is, uh, pretty fitting because we are talking about serious developer tools today. I mean, did you see the Claude Code version two point one point two hundred and nineteen update that just dropped? It is massive. They made claude opus five the default Opus model. And we are talking about a native, under the hood, one million token context window directly in your terminal.

Lachlan Reed

One million tokens. Directly in the command line interface. That is like, I mean, okay, you know I love my old motorbikes, right? Imagine someone dumps a vintage postie bike on your driveway, but it is entirely disassembled. Every single bolt, gasket, and piston ring is just loose in this giant, rusted wooden crate. Before, with smaller context windows, you could only reach in and grab, say, the carburetor and hope it fits the spark plug you pulled out ten minutes ago. Now, with a million tokens, it is like having the entire blueprint, the full history of every mechanic who ever touched it, and every single part laid out on a clean workbench all at once. You can see the whole codebase, start to finish, without the model forgetting where the main function even lives.

James Turner

Exactly! It is the whole garage in your head. But, okay, let us talk about the actual math of this because a million tokens is not cheap if you are slamming the API constantly. They introduced this new, uh, fast mode pricing model for opus five. It is what they are calling ten, ten, ten, fifty per million tokens. So, to break that down, that is ten dollars per million tokens for input, ten dollars for cache writes, ten dollars for cache reads, and then fifty dollars per million tokens for output.

Lachlan Reed

Wait, so ten dollars for input, write, and read, and then fifty for output? That actually makes prompt caching the absolute hero of this setup. Because, think about it. If you are running an AI agent, and it has to scan your entire monorepo on every single turn, you would burn through your API budget before you even finished writing a single unit test. But because the cache read rate is a flat ten dollars per million, once that code is in the cache, those follow up prompts, those little refactoring loops, they are incredibly cheap. It actually makes deep, iterative agent work financially doable.

James Turner

Yes! You get it. The prompt caching is what makes a million token context actually usable for daily coding instead of just a neat party trick. And if you want to set this up right now, the CLI commands are super straightforward. You literally just type slash model or slash config model equals claude opus five right in your terminal. And boom, you are on the new hotness. They also updated the slash fast mode, so now it exclusively routes to Opus five and Opus four point eight. And they actually completely removed the older Opus four point seven from the fast rotation.

Lachlan Reed

Ah, rest in peace, four point seven. You served us well. But, honestly, this is great because if you have custom agent scripts running, the claude api skill has been updated to automatically default to Opus five now anyway. So you do not even have to go in and manually rewrite all your custom tooling. It just, it just works out of the box.

James Turner

Right. It is seamless. And speaking of things just working, there are some really nice quality of life updates in this two point one point two hundred and nineteen release. One that I am super happy about is sandbox dot network dot strict allowlist. It is a new configuration that silently drops any unauthorized external network requests. So instead of the agent constantly pausing and throwing those annoying permission prompts because some package tried to ping home, it just quietly blocks it and keeps running.

Lachlan Reed

Oh, that is huge for local development security. No more prompt fatigue. They also added a DirectoryAdded hook, which is going to be brilliant for anyone working with massive monorepos where directories are getting spun up dynamically. And, oh, my absolute favorite minor fix. If you use Vim mode in the terminal, and you are sitting on an empty normal mode prompt, pressing the left arrow key now takes you straight back into the active agent view. It is such a tiny thing, but as a Vim user, not getting stuck in that terminal limbo is a massive win.

James Turner

It is the little details that keep us sane. Well, that is the quick run down on Claude Code and Opus five. Happy coding, everyone, and we will catch you in the next one.

Lachlan Reed

Too easy. Catch you later, mate.