A free guide by New Money School
Model Cheat Sheet
Most people use the same Claude for everything — burning credits on simple stuff, or getting surface-level answers on the hard stuff. There are 3 Claudes. Each one is built for different work. Here's the cheat sheet.
The Quick Version
Three models. One switch. Better outputs and lower credit burn the second you start using them right.
Haiku 4.5
Fast & lightweight. The credit-saver.
Sonnet 4.6
The daily driver. Anthropic's recommended default.
Opus 4.7
Deep reasoning specialist. The big brain.
How to Switch Models
In claude.ai, click the model name at the top of any chat and pick the model you want. Takes 3 seconds. Available on Free, Pro, and Max plans (Pro and Max get more access to Opus).
The DefaultSonnet 4.6 — Use For 80% of Your Work
Sonnet is the daily driver. Strong reasoning, fast enough for real-time work, capable enough that most problems won't outgrow it. Anthropic specifically recommends Sonnet when you're not sure which model to use.
- Writing emails, reports, drafts
- Building docs, sheets, slides, presentations
- Multi-step workflows where you need it to think and act
- Normal coding tasks
- Research that needs reasoning (not just retrieval)
- Customer support, summarization, content creation
- Anything you'd ask a smart coworker
Example prompts
"Draft a follow-up email to [client] referencing our Tuesday call. Match my normal tone — concise, warm, action-oriented."``"Read this 12-page proposal and tell me which sections need work. Be specific."``"Build me a Q3 launch plan in Sheets — phases, owners, deadlines, dependencies."
The Deep ThinkerOpus 4.7 — When Sonnet Feels Surface-Level
Opus is the reasoning specialist — built for problems that genuinely need deep thinking over multiple steps. Use it when accuracy matters more than speed, or when Sonnet's first answer feels too shallow.
- Long research projects with many sources
- Complex proposals, strategies, or business plans
- Multi-file Cowork sessions where Claude needs to hold a lot in its head
- Code refactors that span multiple files or systems
- Decisions where you need a real second opinion, not a summary
- Anything you'd hire a senior consultant to think through
Pro Tip — Turn On Extended Thinking
Extended Thinking lets Opus take more time to plan and reason before responding. It's now adaptive — easy questions don't waste credits, but hard ones get the full treatment. Worth turning on by default for Opus.
How to enable: Click "Search and tools" in the lower left of your chat → toggle "Extended thinking" on. (Toggling on starts a new chat.)
Example prompts
"Read all 25 articles in /Research. Write a 4-page brief that synthesizes the findings, flags conflicts, and recommends 3 actions."``"I'm choosing between [Strategy A] and [Strategy B]. Pressure-test both. Tell me which one survives scrutiny and why."``"Refactor this 800-line component into smaller, testable pieces. Plan the approach first, then walk me through the changes."
Speed ModeHaiku 4.5 — The Credit Saver
Haiku is fast and lightweight. Most efficient with rate limits. Use it for the high-volume, low-complexity stuff that doesn't need a heavy model thinking it through.
- Quick fact lookups and definitions
- Simple summaries (an email, a paragraph, a meeting note)
- Classifying or labeling things in bulk
- Fast Q&A and one-shot answers
- Extracting specific info from a document
- Anything you'd Google — but smarter
Example prompts
"Summarize this email thread in 3 sentences. Just the summary."``"Categorize these 50 customer feedback notes into 5 themes. Return as a list."``"Quick — what's the standard markup for [industry term] right now?"
Bonus5 Habits That Save Credits On Every Model
These work whether you're using Haiku, Sonnet, or Opus. Five tiny habits that compound fast.
01
Edit your last message — don't send a follow-up
If Claude's answer is off, hover over your previous message and click the edit pencil. Editing forks the conversation, so the wrong answer doesn't carry forward as context. Way more efficient than typing "no, try again."
02
Batch tasks into one message
Three tasks in one message uses fewer credits than three separate messages. Pile them on with a quick numbered list and let Claude work through them.
03
End prompts with "No commentary. Just the output."
Strips Claude's natural preamble and explanation. You get the thing you asked for, nothing else. Cuts response length (and credits) by 30-50% on simple tasks.
04
New topic = new chat. Always.
Long chats accumulate context that gets re-processed with every message. When you switch topics, hit "New chat." Costs nothing. Saves a lot.
05
Long PDF? Convert it to .md before uploading
PDFs work natively in Claude, but they carry formatting overhead that uses extra tokens. For dense or long PDFs, paste the text into a .md file (or just copy/paste the text directly into chat). Same content, fewer credits.
SetupHow to Switch Models in Claude.ai
Step 1: Open any chat at claude.ai. Look at the top of the chat — you'll see the current model name (e.g., "Claude Sonnet 4.6").
Step 2: Click the model name. A dropdown opens with every available model on your plan.
Step 3: Pick the model you want. The chat continues with the new model from your next message forward.
Bonus — Enable Extended Thinking: With a Claude 4 model selected (Sonnet or Opus), click "Search and tools" in the lower left of the chat input area, then toggle "Extended thinking" on. Toggling on starts a new chat.
Plan Notes
Free plan: Sonnet 4.6 + Haiku 4.5. Pro plan: All three models, with limited Opus access. Max plan: All three with the highest Opus limits. Extended Thinking is available on every paid plan that supports Sonnet/Opus.
Three Claudes. One switch. Way better outputs the second you start picking right.