> ## Content Index
> Fetch the complete content index at: https://www.livain.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Fable 5 is back. My terminal still hands me the junior.
- URL: https://www.livain.com/blog/fable-5-is-back-my-terminal-still-hands-me-the-junior/
- Published: 2026-07-06T07:00:00.000Z
- Updated: 2026-08-29T07:59:41.000Z
- Description: Fable 5 returned on 1 July — but Claude Code still defaults to Opus 4.8. A senior partner in one window, a junior developer in the other. Know which one you're talking to.
- Author: Remco Livain
- Tags: AI, Claude Code, Marketing Tech

On the first of July, [Claude Fable 5](https://www.anthropic.com/news/claude-fable-5-mythos-5?ref=livain.com) came back. If you missed the saga: Anthropic's first Mythos-class model launched in early June, dazzled everyone for three days, and then [disappeared behind a US export-control order](https://venturebeat.com/technology/anthropic-is-bringing-back-claude-fable-5-globally-after-us-lifts-export-control-order-where-can-enterprises-access-it?ref=livain.com) after researchers demonstrated a jailbreak. It returned this week with [a new safety classifier](https://thehackernews.com/2026/07/anthropic-restores-claude-fable-5-after.html?ref=livain.com) and, for now, rationed usage.

I noticed its absence more than I expected to. Three weeks ago I wrote that [Fable had us dazzled while Opus 4.8 quietly did my actual work](https://www.livain.com/blog/opus-4-8-quietly-doing-the-work/) — and that was true. But the weeks without it taught me the difference in a way the launch week didn't. Strategy sessions got a little flatter. The second opinion got a little less second.

### The senior is back — in one room

Here's the part nobody puts in the announcement: Fable 5 is back in chat and in Cowork, but in Claude Code — the terminal, where code actually gets written and shipped — [the default model is still Opus 4.8](https://code.claude.com/docs/en/model-config?ref=livain.com). You can switch manually, and with the current usage limits you'll think twice before doing it for routine work.

So my week feels split in two. In the strategy room I have the senior partner again — the one who pushes back, catches the flaw in my framing, thinks past the brief. Then I walk over to the terminal and pair with what I can best describe as a very good junior software developer: fast, tireless, wide-ranging, and in need of a spec and a review before anything it builds touches production.

> The upgrade isn't that the model got smarter. It's that I stopped assuming the same assistant sits behind every window.

### Defaults are product decisions, not recommendations

It's easy to read a default as advice — "this is the right model for this job." It isn't. Defaults balance cost, capacity and safety across millions of users; Fable costs roughly double Opus per token, and Anthropic is clearly managing demand after the restart. Reasonable choices — for them. Whether they're right *for your task* is your call, and most people never make it, the same way [most people never look at what the thinking dials actually do](https://www.livain.com/blog/ultracode-vs-effort-high-claude-code/).

My working split, one week in: judgement-heavy work — positioning, architecture decisions, anything where a wrong frame costs weeks — goes to Fable deliberately. Execution — the well-specified build, the refactor, the data pull — stays with Opus, which remains excellent and half the price. The trap is the middle: agentic coding sessions that *feel* like execution but quietly contain architecture decisions. That's where the junior needs the senior in the room, and where I now switch models on purpose instead of by default. It's the same sequencing lesson as [designing the app in Claude before letting Claude Code build it](https://www.livain.com/blog/i-designed-an-app-in-claude-before-building-it/): put the strongest judgement at the decision, not at the keyboard.

### A better assistant still isn't the leverage

The uncomfortable, familiar conclusion: the model tier matters less than knowing which decisions deserve it. A novice with Fable 5 gets confident-sounding output faster. An operator who knows [which levers are the right ones](https://www.livain.com/blog/the-death-of-generic-ai-why-deep-domain-expertise-is-the-only-real-leverage-left/) gets leverage from either model — and knows when the junior is enough.

Glad you're back, Fable. Now check who's actually sitting at your terminal.

### Sources & further reading

**External**

- [Anthropic: Claude Fable 5 and Claude Mythos 5](https://www.anthropic.com/news/claude-fable-5-mythos-5?ref=livain.com)
- [VentureBeat: Anthropic brings back Claude Fable 5 globally](https://venturebeat.com/technology/anthropic-is-bringing-back-claude-fable-5-globally-after-us-lifts-export-control-order-where-can-enterprises-access-it?ref=livain.com)
- [The Hacker News: Fable 5 restored with new jailbreak classifier](https://thehackernews.com/2026/07/anthropic-restores-claude-fable-5-after.html?ref=livain.com)
- [Claude Code docs: Model configuration](https://code.claude.com/docs/en/model-config?ref=livain.com)

**Related posts**

- [Fable 5 had us dazzled. But Opus 4.8 is quietly doing my work.](https://www.livain.com/blog/opus-4-8-quietly-doing-the-work/)
- [Ultracode isn't a bigger thinking dial. It's Claude deciding how to split the work.](https://www.livain.com/blog/ultracode-vs-effort-high-claude-code/)
- [I designed an app in Claude before building it. Claude Code shipped it in a day.](https://www.livain.com/blog/i-designed-an-app-in-claude-before-building-it/)
- [The death of generic AI: why deep domain expertise is the only real leverage left](https://www.livain.com/blog/the-death-of-generic-ai-why-deep-domain-expertise-is-the-only-real-leverage-left/)