Fable 5, Opus, or Sonnet: which Claude for client work

Revised October 10, 2026. Updated for the September releases: Fable 5.1 replaced Fable 5 at the top of the lineup, and Opus 5.5 replaced Opus 5. The rule and the task table didn't need to change, because they were written to tiers rather than model names. Earlier versions of this article and their notes are kept with its source.

For six weeks this summer, the answer to "which Claude model should I use" kept changing. It settled in late July, and the settled version has a twist worth knowing before you read the rest: which model sits at the top of your picker depends on which plan you pay for. The September releases renamed the models without changing that.

So the picker is a real decision, just not the one it was in June. Most of what's written about "Opus vs Fable" is for developers: benchmark charts, per-token prices, context-window trivia. None of it tells you which model should write Friday's client update. Here's the consultant version: one rule, then defaults, task by task.

What's actually in the picker

Skip the spec sheets. Here's the working translation.

Fable 5.1 is Anthropic's most capable model, the tier above Opus. Anthropic positions it for "demanding reasoning and long-horizon agentic work," which for a consultant means long, multi-step jobs: it holds a whole deliverable together from start to finish instead of drifting in the middle. The trade is time, because hard asks take longer. Whether it's in your picker at all depends on your plan. We covered what the top tier changes for consulting work; this piece is about when to choose it.

Opus 5.5 is the one that matters for most consultants. Anthropic says it "performs at the level of Claude Fable 5.1 on most work," and it's included on every paid plan. Strong judgment, strong writing, and fast enough for interactive work. It also has an adjustable effort setting, which gets its own section below because it's the most useful lever in the picker.

Sonnet 5.5 is the balance of speed and intelligence: noticeably quicker in conversation, and more than capable for work whose shape you already know.

If your picker shows a lighter option such as Haiku, use it for instant answers: lookups, quick transformations, "make this a bullet list." You may rarely choose it on purpose, and that's fine.

Which of these you can reach is the one thing here that depends on what you pay, and it's the only part of this article that names a plan:

As of October 10, 2026. Fable 5.1 is included on Max for up to half your weekly usage. On Pro it isn't included, and reaching it means pay-as-you-go usage credits. Opus 5.5 is on every paid plan. The full table, with Team and Enterprise, is on which model you get on which plan, and Anthropic's plan page is the authority.

Two practical notes before the rule. First, cost: the models in your picker draw on the subscription you already pay for, plans meter that usage, and the bigger models draw it down faster. Spending your top model on a subject line is paying caviar prices for toast. Second, names: Anthropic ships new versions on its own schedule, and this article has been revised several times for exactly that reason. The tiers persist: a top model for the hardest work, a strong default, a fast one. Learn the tiers and the rule below survives every release. That's the narrow version of a broader point we've made about not building your practice on one model.

The rule: match the model to the leash

Forget capability rankings. The question that picks the model is this: how long is the leash on this task, and what does a wrong answer cost you?

Leash length is how much work you hand over between check-ins. A subject-line rewrite is a ten-second leash: you see the output immediately, a miss costs nothing, and you iterate without thinking about it. "Here's the transcript, my services file and my rate floor; draft the proposal" is a long leash. The model works through many steps before you see anything, and a mistake in the middle compounds quietly until the end.

Put those two tasks on the same model and you've made an error in one direction or the other.

The second half of the rule, what a wrong answer costs, settles the borderline cases. A cheap mistake gets caught for free: you were reading the output anyway, and the fix is one more message. An expensive mistake is the kind that survives your review because the document looks finished, like a proposal with a plausible but wrong price, or a case study quoting something the client never said. The polish of a long output hides problems. The more finished a draft looks, the stronger the model you want to have made it, and the harder you should read it before it goes out.

Short leash, low stakes: the fast model. You're in a tight loop, waiting is the main cost, and you check every output yourself anyway. Extra depth has nowhere to help.

Medium leash, judgment at each step: the strong model, steered. The work happens in stages and your judgment gates each one: scope before price, price before draft. You want the sharpest output at each stage and a loop quick enough that you don't lose momentum between them.

Long leash, whole deliverable: the strongest model you can reach. Handing over a whole job rewards the model that can hold all of it in view, and the middle of a long run is where weaker models used to drift. This is the one row where your plan decides for you, and where Fable 5.1 earns its place if you have it.

A concrete contrast: the Friday client update is a short-leash task in business clothes. It has a known shape, your inputs, and your proofread before it goes out. The proposal built from a discovery transcript is a long-leash task in the same clothes, with a dozen quiet decisions about scope, sequence and emphasis between your brief and the draft. Same client, same week, opposite ends of the picker.

Task by task

Default assignments for the recurring work of a solo practice. Adjust them to taste; the point is to have defaults instead of re-deciding in front of every blank conversation.

The column that matters is the tier, not the model name. Names change every few months, and this table shouldn't have to.

As of October 2026 those tiers read: Sonnet 5.5 for fast, Opus 5.5 for the strong default, and Fable 5.1 for the top tier if your plan includes it. Names will change again, so check your own picker rather than trusting the ones printed in any article, including this one. When they do change, which model you get on which plan is the page we update first.

The pattern is simple. Routine work goes to the fast tier, judgment-steered work goes to the strong default, and anything handed over whole goes to the top tier you can reach, which is the one row where your plan decides for you. When a task doesn't fit a row, ask the leash question and it sorts itself out.

The effort setting maps onto the leash rule exactly. The current models let you choose how much work they do before answering, from Low up to Max. Short leash, low stakes: leave it low and enjoy the speed. Long leash, expensive mistakes: turn it up and let the model check itself. In practice this narrows the gap between the strong default and the top model for most consulting work. A well-briefed Opus 5.5 run at high effort covers most of the deliverable work that used to call for the tier above.

The Friday-update default surprises people. Surely the better model writes the better update? Marginally, maybe. But a good weekly update comes from its shape and its inputs, not from model depth, and you proofread it regardless. Save the depth for work where mistakes are expensive.

The proposal row is the one that moved this year. A year ago the proposal was a steered task: you chained the stages because no model could be trusted with the whole thing at once. With models built for long runs, it becomes a delegated task for consultants who can write a complete brief covering the objective, inputs, constraints and what done looks like. If your handoffs aren't that complete yet, keep steering. A long leash on a vague brief just produces a longer wrong document, and no effort setting fixes a missing rate floor.

The discovery debrief deserves its own note, because it's the highest-stakes steered task on the list: what you take from that call feeds the scope, the price and the proposal that follow. If you want a tested starting point rather than a blank page, our free Discovery Debrief is a single-prompt workflow with a Claude skill and a worked example. It runs the same whichever model you pick.

When the top model is the wrong choice

Reaching for the most capable model deserves a section of honest no.

When you need the loop, not the leash. Interactive drafting, where you react line by line and talk a document into shape, wants the fastest model that's good enough. Deeper reasoning shows up as waiting, and waiting breaks the rhythm of steered work. A brilliant answer ninety seconds late is a worse partner than a very good answer now. This is also what the effort setting is for: turn it down rather than switching models.

When the shape is known. A weekly update, a meeting recap, an invoice reminder: structure plus your inputs solves the task. Extra capability has nowhere to go, so you pay its costs in time and usage for nothing.

When you're rationing usage. Heavy top-tier runs draw down plan allowances faster than the smaller models do, and where the top tier is included at all it's capped rather than unlimited. If you're bumping into limits by the end of the month, check whether routine tasks have drifted up to the bigger model out of habit.

When it isn't in your plan. If reaching the top tier means buying usage credits, every long run is a cost you chose, and that's a perfectly workable way to operate. Keep the model your plan includes as your default and spend deliberately on the one deliverable a month that warrants it. What you shouldn't do is treat a credit balance as a general upgrade: the leash rule still puts most of your week on the fast tier and the strong default. Whether the top tier is worth changing plans for comes down to your billing rate, not a benchmark, and the plan page we keep walks through that math.

One footnote so it doesn't surprise you: the top-tier models ship with safety classifiers around a handful of research-heavy topics, such as security research and biology. When Fable 5 launched, Anthropic said fewer than five percent of sessions encountered them, and consulting work rarely goes near those areas. If a conversation ever behaves oddly, the companion piece covers the practical response: start a fresh conversation, ask more plainly, and move on.

Set your defaults this week

Three moves now and one for the calendar. None takes longer than a coffee.

Write the defaults down. List your five most frequent Claude tasks. Give each a tier from the table above, then write today's model name next to it, and put the list somewhere you'll see it. The value isn't in the assignments themselves; it's in never re-deciding at a blank conversation again.

Run one A/B test. Take a deliverable you produce regularly. Same inputs, same brief, two models: your current habit and the table's recommendation. Read both drafts cold and pick the one you'd actually send. That's evidence from your own practice, not a stranger's benchmark screenshot. If you want a ready-made deliverable to run it on, our free Friday Update Brief is a complete handoff with a sample week of notes and a reference output. It costs an email address, and because both runs work from identical inputs, the comparison is fair.

Move the routine work down. If everything has been running on the biggest available model out of habit, move the known-shape tasks to the fast tier for a week. You'll notice the speed straight away, you probably won't notice a drop in quality, and the usage headroom comes back for the long runs that need it.

And the calendar one: put an expiry date on your defaults. Models change under you, and assignments go stale without announcing it. The names in the table above changed twice between June and October. When the next release lands, re-run the A/B on your highest-stakes task before believing anyone's write-up, including this one. It takes ten minutes.

The deeper question underneath all of this (which work to hand over at all, and which stays yours however capable the models get) doesn't change with the picker. That's the operating frame of The Solo Consultant's AI Playbook, and it's deliberately model-agnostic. Pick the model per task. Keep the judgment on retainer.