Claude Model Family Basics

7 examples to get you started with the Claude Model Family - 5 basic and 2 intermediate.

Prerequisites

• A Claude.ai account (Free, Pro, or higher) or API access through the Claude Console.
• No installation needed for Claude.ai - the model picker is built into the chat interface.
• For API use, you'll need an API key and the model name string for whichever tier you choose.

Basic Examples

1. Finding the Model Picker

The model picker sits at the top of the Claude.ai chat window, above the message box.

Clicking it opens a dropdown listing the available tiers - typically Claude Sonnet 5 (marked as default), Claude Opus 4.8, Claude Fable 5, and Claude Haiku 4.5, depending on your plan.

• Free and Pro plans default to Claude Sonnet 5 automatically.
• Higher tiers (Opus 4.8, Fable 5) may require a paid plan or usage credits.
• You can switch models mid-conversation; earlier messages stay in context.
• Haiku 4.5 is often available even on lighter plans, since it's the cheapest tier to serve.

Related: Why Anthropic Offers Four Different Claude Models - the reasoning behind the tiers you're picking from

2. Starting with the Default: Claude Sonnet 5

For most everyday questions, the right move is simply to not think about the model picker at all.

Example prompt: "Summarize this article in three bullet points and suggest a headline."

• Claude Sonnet 5 became the default model for Free and Pro plans at its June 30, 2026 launch.
• It's tuned for balanced, everyday use - strong at coding, writing, and agentic tasks.
• Intro pricing on the API is roughly $2 input / $10 output per MTok through August 31, 2026.
• If a task feels routine, Sonnet 5 is very likely the right tier without any manual switching.

3. Escalating to Claude Opus 4.8 for Hard Reasoning

Some tasks are dense enough that it's worth manually switching up a tier.

Example prompt: "Walk through the second-order effects of this pricing change on our three customer segments, and flag any assumptions you're uncertain about."

• Opus 4.8 is Anthropic's flagship reasoning model, built for harder, more ambiguous problems.
• It ships with a 1M token context window by default - useful for reasoning across long inputs.
• It costs roughly $5 input / $25 output per MTok, noticeably more than Sonnet 5.
• Reach for it when a task needs careful multi-step reasoning, not just fast recall.

4. Reaching for Claude Fable 5 on the Hardest Problems

Fable 5 is the tier to pick when a task is both large and genuinely difficult.

Example prompt: "Review this entire 400-page technical spec, cross-reference it against the attached codebase notes, and identify every inconsistency."

• Fable 5 is Anthropic's top, Mythos-class tier - the most capable widely released model as of ~June 2026.
• It offers a 1M token context window and a 128K token maximum output, the largest output ceiling in the lineup.
• Its always-on adaptive thinking automatically applies more internal reasoning to harder queries, without you requesting it.
• It's also the most expensive tier, at roughly $10 input / $50 output per MTok - reserve it for work that truly needs the ceiling.

5. Routing Quick, High-Volume Work to Claude Haiku 4.5

Not every task deserves a heavyweight model, even if you could afford one.

Example prompt: "Classify this list of 50 support tickets as billing, technical, or general - one word each."

• Haiku 4.5 is the fastest and cheapest tier, priced around $1 input / $5 output per MTok.
• It carries a 200K token context window - smaller than Opus or Fable, but plenty for short, focused tasks.
• It's built for high-volume, latency-sensitive work: quick classification, short replies, simple extraction.
• Using Haiku for simple, repetitive tasks keeps costs down without sacrificing quality where it doesn't matter.

Intermediate Examples

6. Matching a Multi-Step Project to a Mix of Tiers

Real projects rarely fit one model tier end to end.

Example scenario: drafting a quarterly report from raw data.

1. Use Haiku 4.5 to clean and tag hundreds of raw survey responses quickly and cheaply.
2. Use Sonnet 5 to draft the report narrative and pull in the tagged data - everyday writing and synthesis work.
3. Use Opus 4.8 to stress-test the report's conclusions and check for reasoning gaps before it goes to leadership.

• Mixing tiers within one project is normal and often the most cost-effective approach.
• Cheap, high-volume steps go to Haiku; synthesis goes to Sonnet; hard verification goes to Opus or Fable.
• This mirrors how the tiers are actually priced - reserving the expensive tiers for the steps that need them.
• Revisit the mix if a "simple" step keeps producing weak results - it may actually need a higher tier.

Related: Choosing the Right Claude Model for Your Team's Task - a fuller decision list for matching tasks to tiers

7. Budgeting Across Tiers for a Team

Once a team relies on Claude for more than occasional chat, budget planning becomes part of picking a model.

Example scenario: a small team estimating monthly API spend before rolling out an internal tool.

1. Estimate rough monthly token volume for the tool's main workflow.
2. Check current per-MTok input/output pricing for each candidate tier (Haiku 4.5 ≈ $1/$5, Sonnet 5 ≈ $2/$10 intro pricing, Opus 4.8 ≈ $5/$25, Fable 5 ≈ $10/$50).
3. Default the workflow to Sonnet 5, and prototype with Haiku 4.5 first if the workflow is high-volume and simple.
4. Reserve Opus 4.8 or Fable 5 for a narrow, clearly-justified subset of calls, not the whole workflow.

• Output tokens cost roughly 4-5x input tokens on every tier, so response length matters as much as prompt length.
• Sonnet 5's intro pricing window (through August 31, 2026) is a good moment to prototype cheaply before rates shift.
• Small pilots on Haiku 4.5 can validate a workflow before committing budget to a pricier tier.
• Pricing and exact rates change - always confirm current numbers before finalizing a budget.

Related: Claude Model Pricing Compared: Fable, Opus, Sonnet, and Haiku - the full side-by-side pricing table

Stack versions: Written against the Claude model lineup current as of ~June 2026 - Claude Fable 5, Claude Opus 4.8, Claude Sonnet 5 (the default), and Claude Haiku 4.5. Model names, pricing, and product features move quickly - verify current specifics at platform.claude.com/docs before relying on them.