Picking the wrong Claude model is like grabbing a chainsaw to slice a sandwich. Overkill, messy, and it drains your budget before lunch. I used to reach for the heaviest model on every tiny task, then stare at my token count wondering where it all went.
Then I came across this sharp little decision tree from a LinkedIn creator who maps every Claude model to the exact job it was built for. The author lays out all four current models and hands you a simple path to route any task to the right one. I think it’s one of the cleanest cheat sheets I’ve seen for this, and it fixes the exact mistake I kept making.
Here are the four models the expert is working with:
- Haiku 4.5: the lightweight token-saver.
- Sonnet 5: the fast, everyday workhorse.
- Opus 4.8: the deep-work model.
- Fable 5: the smartest, most ambitious model.
🧭 The one question that starts everything
The original poster says the whole tree begins with a single check, and this is the part most people skip. The steps below follow the creator’s logic exactly.
- Ask if your task needs a complex answer. If no, stay light with Haiku 4.5 or Sonnet 5. If yes, step up to Opus 4.8 or Fable 5.
- For quick tasks, decide between raw speed (Haiku 4.5) or everyday ease (Sonnet 5).
- For your hardest, most ambitious work, choose deep execution (Opus 4.8) or the smartest reasoning (Fable 5).
That first fork alone saves you from overpaying on simple work. Now let’s walk each branch the way the expert broke it down.
⚡ Haiku 4.5: when speed and low cost win
This LinkedIn creator points to Haiku 4.5 when you need maximum speed and minimal tokens. It’s the featherweight of the family. The author’s practical routine looks like this:
- Chat without files attached.
- Turn on web search.
- Plan in Chat, then build in Cowork.
Why it matters: you save tokens by matching a small model to a small job instead of firing up something ten times heavier. Here’s the example prompt the creator shared for Haiku:
“I want [desired result] with [constraints]. Ask me questions using AskUserQuestion before you start.”
🚀 Sonnet 5: your everyday default
When the task is quick but you don’t need bare-bones speed, the expert reaches for Sonnet 5. It’s the fast, reliable model for simple daily tasks. One tip the author highlights is connecting your apps through Connectors: Slack, Google Drive, Notion, Figma, Granola, Gamma, and more than 50 others.
Why it matters: Sonnet handles the bulk of normal work without the cost of a heavy model, and plugging in your tools means less copy-pasting. The prompt the creator recommends:
“You are a [role]. [Task] this [input]. Keep it under [length]. Tone: [casual/formal]. No preamble – just the output.”
🧠 Opus 4.8: the deep-work model
Now the tree climbs. For hard tasks that don’t quite reach your most ambitious tier, the original poster routes to Opus 4.8. This is the model for multi-step work, agents, and long-thinking sessions. The author’s rules of thumb:
- Use Cowork.
- Set Effort to High. Always.
- Use Skills in Projects, like a writing skill or a formatting skill.
Why it matters: Opus is built to grind through complex, layered problems, and running it at high effort inside a project keeps your context organized. The example prompt from the creator:
“/[skill] topic: [topic]. DO NOT start yet. Ask me clarifying questions (use AskUserQuestion) so we can refine the approach step by step.”
The expert also shares a smart wrap-up move after a Cowork session:
- Download your file.
- Or convert it into a Claude Skill.
- Start a fresh session to save tokens.
🌟 Fable 5: the top of the tree
At the very top sits Fable 5, described by the author as the smartest model Claude offers. This is the one for deep research and heavy analytical decisions. The prompt the creator suggests leans into that reasoning power:
“Here is my goal: [goal]. Here are my constraints: [constraints]. Think through the tradeoffs before answering, propose 2–3 approaches, and recommend one with your reasoning.”
⚠️ A serious warning about Fable 5
Here’s where I really appreciated the creator’s honesty. Fable is powerful, but the author is blunt about using it carefully:
- It costs extra, moving to pay-per-use.
- Only about 10% of tasks actually need it.
- Use it for one or two turns of strategy, then switch back down to Opus.
- Long conversations get expensive, because Claude re-reads the whole thread every single turn.
The rule the expert lands on: when Opus gets stuck, escalate to Fable. Everything else, route down the tree.
Why this cheat sheet is worth saving
What I love about the way this contributor framed it is the mindset shift. You stop treating model choice as a habit and start treating it as a decision. Match the model to the task, keep your heaviest firepower for the moments that truly need it, and your token spend stays sane.
Route down whenever you can. Escalate up only when you’re genuinely stuck. That simple discipline is what stops teams from burning tokens on autopilot. ♻️
This is just the highlight reel of the author’s full breakdown, including the exact routing logic step by step. Check out the original LinkedIn post for the complete decision tree and every detail from the creator.