Claude Model Selection Guide: Fable, Opus, Sonnet, and Haiku for Business Owners
Which model should you use, what does the effort slider do, and how does it all affect your bill
Who this article is for: This is for power users, people managing costs, anyone who has hit usage limits, or business owners who want to understand what is happening under the hood. If you are new to Claude, you do not need to read this first — start with the Commands and Tips guide or the Growth Mindset article and come back here when you are ready to optimize.
A Note on When This Actually Matters
Here is the honest truth: most people do not need to think about model selection right away. Claude works well out of the box with its default settings.
Where this becomes important:
- You are hitting usage limits and need to stretch your plan further
- You are managing costs across a team or organization
- You are building AI into your product — this is where model selection is critical, because you are paying per customer interaction and need the best quality at the best price
- You want to optimize your personal workflow for speed and cost
What I have observed advising teams: when engineers are not paying the bill, they default to the most powerful model for everything and burn through tokens fast. And honestly, it is annoying to constantly think about which model to use — it can feel like it slows you down. So here is my recommendation: know that you have the options, optimize when you choose to, and do not let it paralyze your workflow.
For individual users on a Pro plan, the defaults are fine until they are not. For teams and businesses building AI features, this guide is essential reading.
Why Model Selection Matters for Your Bottom Line
The Consultant Analogy
Think of Claude's models like hiring from a consulting firm:
| Model | Who They Are | What They Cost | When You Call Them |
|---|---|---|---|
| Fable | A specialist brought in for a once-in-a-decade project | $$$$ | Cutting-edge research, frontier capability — most businesses will never need this |
| Opus | The senior partner | $$$ | Complex strategy, high-stakes decisions, nuanced analysis |
| Sonnet | An experienced associate | $$ | Day-to-day work, reliable execution, solid judgment |
| Haiku | A sharp junior associate | $ | Quick tasks, simple lookups, high-volume repetitive work |
You would not pay senior partner rates to have someone rename a file. And you would not send a junior associate to negotiate a critical deal. The same logic applies here. And you definitely would not bring in an outside specialist at a premium rate to handle work your senior partner can do just as well.
What You Are Actually Paying For
Every time you send a message to Claude, you are paying for two things:
- Input — the text you send (your question, any documents you attach)
- Output — the text Claude generates in response
Bigger models think harder and produce higher-quality output, but they cost more per word in both directions. The model you choose is the single biggest lever on your bill.
The Four Claude Models Explained
Claude Fable — The New Specialist (Skip This for Now)
Fable is Anthropic's newest and most expensive model. It was released in 2025 and sits above Opus in both capability and cost.
The honest take: most businesses have no reason to use Fable.
Fable is designed for frontier use cases — advanced research, highly complex multi-step reasoning, problems that are genuinely at the edge of what AI can currently do. If that sentence does not describe your work, Fable is not for you.
The cost premium is significant. You are paying materially more than Opus for capabilities that only matter in a narrow range of specialized scenarios. For typical business tasks — drafting, analysis, summarization, strategy — Opus already handles these at a level that is hard to improve on. Fable does not make your emails better. It does not improve your meeting summaries. The gap between Opus and Fable only shows up in work that most businesses simply are not doing.
Who actually needs Fable:
- Researchers working on technical problems at the frontier of their field
- Teams building specialized AI systems where incremental capability gains translate to real business outcomes
- Use cases where Opus has genuinely hit a ceiling and you have tested that conclusion
Who does not need Fable:
- Anyone using Claude for standard business operations
- Teams using Claude Code for productivity and workflows
- Anyone who has not already exhausted what Opus can do
If you are reading this guide to understand which Claude model to use for your business, Fable is almost certainly not the answer. Stick with the Opus/Sonnet/Haiku framework below and revisit Fable only if you have a specific, validated reason.
Claude Opus — Your Senior Partner
Opus is Anthropic's most capable model. It reasons deeply, catches subtleties other models miss, and handles multi-step problems that require holding a lot of context in its head at once.
Use Opus when:
- You are drafting a strategy memo that needs to account for multiple competing factors
- You are reviewing a contract and need to catch implications buried in the fine print
- You are debugging a problem where the obvious answer is not working
- You need Claude to reason through a decision with real trade-offs
- The output will be seen by clients, investors, or your board
Do not use Opus when:
- You need a quick summary of a document
- You are reformatting data from one shape to another
- The task has a clear, simple answer
- You are doing high-volume repetitive work
Real-world example: You paste in three competing vendor proposals and ask Claude to compare them across cost, capability, risk, and long-term fit. Opus will catch nuances in the pricing structures and flag terms that could become problems at scale.
Claude Sonnet — Your Experienced Associate
Sonnet is the workhorse. It is fast, capable, and costs a fraction of Opus. For the majority of business tasks, Sonnet delivers quality that is indistinguishable from Opus at roughly one-fifth the price.
Use Sonnet when:
- You are drafting emails, internal documents, or reports
- You are summarizing meeting notes or long documents
- You are brainstorming ideas or outlining a project
- You need to analyze data and pull out key insights
- You are doing most of your daily work with Claude
This should be your default model. Start every session on Sonnet. Switch to Opus only when you hit a task that genuinely needs deeper reasoning.
Real-world example: You paste in your weekly sales data and ask Claude to identify trends, flag anomalies, and draft a summary for your team standup. Sonnet handles this quickly and accurately.
Claude Haiku — Your Junior Associate
Haiku is the fastest and cheapest model. It is ideal for simple, well-defined tasks where speed and volume matter more than depth.
Use Haiku when:
- You need to classify or categorize a large batch of items
- You are extracting specific data points from structured documents
- You need quick yes/no or short-answer responses
- You are doing repetitive formatting tasks
- You are processing high volumes where cost adds up fast
Do not use Haiku when:
- The task requires judgment or nuance
- You need the output to be polished enough for external audiences
- You are asking open-ended strategic questions
Real-world example: You have 200 customer support tickets and need them categorized by topic (billing, technical, feature request, complaint). Haiku does this quickly and cheaply.
The Effort Slider: Low, Medium, and High
The effort slider is separate from the model. It controls how long the model thinks before it responds — not which model you are using.
What It Actually Controls
Think of it as asking a consultant to review a document:
| Effort Level | What Claude Does | Analogy |
|---|---|---|
| Low | Quick scan, surface-level response | Glancing at the first page and giving you a gut reaction |
| Medium | Careful read, balanced analysis | Reading the full document and highlighting the key points |
| High | Deep analysis, considers edge cases and implications | Studying it with a highlighter, cross-referencing other documents, and writing up a thorough assessment |
When to Dial Up vs Dial Down
Use Low effort for:
- Questions with straightforward answers
- Quick reformatting or rephrasing tasks
- When you just need a starting point and plan to iterate
Use Medium effort for:
- Most everyday work (this should be your default)
- Drafting content, summarizing documents, analyzing data
- Tasks where you want solid quality without waiting longer
Use High effort for:
- Important decisions where missing a detail could be costly
- Complex analysis with multiple variables
- Anything going to an external audience where quality is critical
- Tasks where you have already tried Medium and the output was not thorough enough
How Models and Effort Work Together
Model selection is who does the thinking. Effort is how hard they think. Combining them lets you fine-tune cost and quality for every task.
| Task | Model | Effort | Why This Combination |
|---|---|---|---|
| Renaming a batch of files | Haiku | Low | Trivial task. Save your money. |
| Drafting a team update email | Sonnet | Low | Standard writing, you will edit it anyway |
| Summarizing a 30-page report | Sonnet | Medium | Needs thoroughness but not Opus-level depth |
| Analyzing competitor pricing strategy | Opus | Medium | Complex analysis, but Medium effort is sufficient for a first pass |
| Reviewing a partnership agreement | Opus | High | High stakes, you need every detail caught |
| Categorizing 500 survey responses | Haiku | Low | Volume task, simple classification |
| Writing an investor update | Opus | High | External-facing, nuanced, high-stakes |
| Generating interview questions | Sonnet | Medium | Creative task, solid quality needed |
| Reformatting a CSV into a table | Haiku | Low | Mechanical task, no judgment needed |
Cost Implications: What This Means for Your Monthly Bill
Exact pricing changes, so rather than quoting specific numbers that will become outdated, here is the relative scale that stays true:
The Relative Cost Scale
If a Haiku task at Low effort costs $1, here is what the same task costs across models and effort levels:
| Low Effort | Medium Effort | High Effort | |
|---|---|---|---|
| Haiku | $1 | $2 | $3 |
| Sonnet | $5 | $10 | $15 |
| Opus | $15 | $30 | $50 |
| Fable | $50+ | $100+ | $200+ |
These are not actual prices — they are ratios showing how quickly cost compounds. The jump from Haiku/Low to Opus/High is roughly 50x. Fable compounds that further. For most business tasks, the cost increase delivers no meaningful improvement in output quality.
The 80/20 Rule
For most business users, approximately 80% of tasks should run on Sonnet at Medium effort. This is your everyday default.
The remaining 20% splits between:
- Haiku for volume and speed (batch processing, simple extractions, formatting)
- Opus for depth and stakes (strategy, contracts, anything where missing something costs real money)
If you are running everything on Opus at High effort, you are dramatically overspending. If you are running everything on Haiku, you are leaving quality on the table.
Common Mistakes Business Owners Make
Mistake 1: Using Opus for everything because "I want the best." You are paying 10-15x more than you need to for most tasks. Sonnet handles 80% of business work at indistinguishable quality.
Mistake 2: Never changing the model. The default model is a starting point, not a permanent setting. Switching takes two seconds and can save you significant money over a month.
Mistake 3: Ignoring the effort slider entirely. Most people never touch it. But dropping to Low effort for simple tasks and raising to High for important ones gives you better results and lower costs simultaneously.
Mistake 4: Using Haiku for tasks that require judgment. Haiku is fast and cheap, but it will miss nuances. If the task involves trade-offs, ambiguity, or stakes, step up to Sonnet or Opus.
Mistake 5: Using Fable because it is the newest model. New does not mean better for your use case. Fable is expensive, and for the vast majority of business tasks it delivers no meaningful improvement over Opus. Unless you have a specific, validated reason to need Fable, treat it as out of scope. The cost is not justified for standard business operations.
Quick-Start Recommendation
If you remember nothing else from this guide:
- Set Sonnet as your default. It is the best balance of cost, speed, and quality for business use.
- Drop to Haiku when the task is simple, repetitive, or high-volume.
- Step up to Opus when the stakes are high, the problem is complex, or the output is going to someone important.
- Leave effort on Medium as your default. Drop to Low for quick tasks, raise to High for important ones.
That four-part system will get you 90% of the way to optimal usage without overthinking it.
This article is part of The Complete Guide to Claude AI for Business Owners. Next: Commands, Tips, and Best Practices.