2026年07月10日

Claude Fable 5 in Claude Code — On Your Phone (2026)

Claude Fable 5 is available in Claude Code. Here's how to use it for coding, what its benchmarks and pricing actually mean, and how to run long Fable 5 tasks from your phone with SeaWork — your code stays local.

To use Claude Fable 5 in Claude Code, update Claude Code and select Fable 5 as your model — it’s been available globally since July 1, 2026 on Claude.ai, Claude Code, and Claude Cowork. Below: what its benchmarks and $10/$50 pricing actually mean for coding, three real commands, and how to run a long Fable 5 task from your phone with SeaWork while your code stays on your machine.

That’s the answer if you just needed the command. The rest is the honest version — because Fable 5’s whole personality is long, hard tasks, and a model that grinds for an hour is precisely the one you don’t want to babysit at your desk.

What Fable 5 is, and the timeline worth knowing

Claude Fable 5 is the first model in Anthropic’s Claude 5 family — the general-availability, production-safeguarded deployment of the same weights as the restricted Claude Mythos 5 (Anthropic). If you’ve seen conflicting release dates, here’s the real sequence, and it’s worth being precise about:

June 9, 2026 — Fable 5 and Mythos 5 launched (Anthropic; TechCrunch).
Mid-June — access was suspended after US export controls landed on the newest models (InfoQ).
July 1, 2026 — with the controls lifted, Anthropic redeployed Fable 5 globally on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork, adding a new classifier that reroutes flagged requests to Claude Opus 4.8 (Anthropic; MarkTechPost).

So if you tried Fable 5, lost it, and wondered where it went — that’s why. As of July 1 it’s back and available in Claude Code.

What the benchmarks and price actually mean for coding

The headline number: on SWE-Bench Pro — a test of difficult, real-world engineering tasks — Anthropic reports Fable 5 at 80.3%, against 58.6% for GPT-5.5 (Anthropic; ecorpit). That’s a large gap on hard coding work, and it’s the number that matters most if your day is real repositories rather than toy problems.

Then the price, stated honestly because it’s the catch: $10 / $50 per million tokens (input / output) — roughly double Opus 4.8 and double GPT-5.5 on input (Anthropic; finout). There’s batch pricing at $5 / $25 for work that isn’t latency-sensitive.

Here’s the first-principles read on when that price is worth it. A coding agent’s cost is dominated by the tasks where it flails — burns tokens going in circles, produces a wrong diff you have to catch and revert, or needs three rounds to land something. On a genuinely hard task, a more capable model that gets it right in one pass can be cheaper in total than a cheap model that thrashes, even at double the per-token rate. So Fable 5 earns its price exactly on the hard, long, high-stakes tasks — the cross-cutting refactor, the bug that’s beaten lesser models — and is overkill for boilerplate. Match the model to the task, not to the leaderboard.

One caveat to keep honest: these are the vendor’s own reported numbers on their chosen benchmark. Treat them as a strong signal of ranking, not a guarantee for your codebase. Your repo is the real benchmark.

Why Fable 5 makes remote control more useful, not less

This is the part that’s specific to Fable 5. Its strength is long, sustained, hard tasks — the runs where you hand it something big and it works for a long stretch. That capability creates a very specific problem: the more independently a model can grind, the longer it runs unattended, and the more it costs you when it goes unwatched down a wrong path or freezes on a permission prompt four minutes in.

In other words: the better the model is at long tasks, the more you need to be able to watch and steer those tasks without being chained to your desk. A model you kick off and check on in ten seconds doesn’t need remote control. A model you kick off before lunch and that’s still working when you get back — that one does. Fable 5 is squarely the second kind.

Run Fable 5 in Claude Code

Update Claude Code, then select Fable 5. Check the exact selector against claude --help since model selection may be a flag or a config entry, but the shape is:

# Run a task in Claude Code on Fable 5
claude --model claude-fable-5 "refactor the auth layer and keep the test suite green"

That’s fine at your desk. The moment you start a long Fable 5 run and want to walk away, you want it on a leash you can hold from your phone.

Steer a long Fable 5 task from your phone with SeaWork

The principle is simple: don’t move the agent, move the interface.

Claude Code on your laptop is already wired into your world — your credentials, your MCP servers, your actual files in your actual repo with your actual environment and tool versions. That setup is correct. The clean pattern isn’t to recreate it in a hosted sandbox; it’s to leave the agent where it already works and bring the control surface to your pocket.

That’s what SeaWork does. The SeaWork daemon runs on your development machine and drives the Claude Code agents already there. Your phone connects over an end-to-end encrypted relay that only forwards encrypted packets — it can’t read your prompts, diffs, or logs, because it never holds the keys. Your code, logs, and Claude credentials stay on your machine. Prefer no relay at all? The daemon is yours to self-host on your own network.

You don’t install a separate daemon — the desktop app ships with it. Three steps:

Install the desktop app. The SeaWork daemon is built in, so your dev machine is ready.
Pair your phone. Open the mobile app and scan the pairing code to establish the encrypted link.
Start a Fable 5 Claude Code agent from the phone, the desktop, or the terminal.

The CLI shows what’s happening under the hood. These are real commands:

# Start a Claude Code agent on a long task (running Fable 5 in your setup)
seawork run --provider claude "refactor the auth layer"

# List your agents (running and finished)
seawork ls -a

# Follow an agent's logs live
seawork logs <id> -f

Start something like that, then pick it up on your phone. On a long Fable 5 run this is where it pays off: you watch every tool call as it happens, read the full conversation, and message the agent mid-run to course-correct the moment it drifts — before it spends another twenty minutes and another few dollars going the wrong way. And when it hits an approval prompt — the “allow this command?” gate that would otherwise freeze the run until you’re back at your desk — the request lands on your phone and you approve or reject it in seconds.

One app for Claude Code and Codex

Fable 5’s price also changes how you mix models. You might run a cheaper model for the mechanical bulk of a feature and hand Fable 5 the one genuinely hard subtask — or reach for Codex on something else in the same afternoon. SeaWork runs Claude Code and Codex from a single app, side by side, and in parallel: one agent on the hard refactor, another writing tests, a third drafting migration notes — each in its own workspace, each visible at a glance so you see what’s running, what’s waiting on you, and what finished.

For anyone splitting implementation, tests, and review across separate worktrees, a single terminal scrollback stops scaling fast. A phone that shows you every agent at once fits that far better.

Where to start

Using Fable 5 in Claude Code is the easy part — update, select the model, go. The harder, more valuable part is not being pinned to your desk while a long, expensive Fable 5 run grinds through a hard task. That’s exactly the case where watching and steering from your phone turns wasted minutes and wasted tokens into a two-second course-correction.

SeaWork is free. Download SeaWork, read more about Claude Code on SeaWork, or see how to use Claude Code from your phone for the wider survey. For steering a single session the first-party way, Claude Code remote control covers /rc.