Claude Models Explained: Fable, Opus, Sonnet and Haiku

Claude comes in four current model lines: Fable, Opus, Sonnet and Haiku. As of October 24, 2026 they are Claude Fable 5.1 (most capable), Opus 5.5 (agentic coding and knowledge work), Sonnet 5.5 (speed plus intelligence) and Haiku 5.5 (fastest and cheapest). All four have a 1M-token context window and 128K max output.

This page explains what each line is for, lists the exact API model IDs, and shows how to choose. Every spec and price below comes from Anthropic’s models overview and pricing page, checked October 24, 2026. Models change several times a year, so confirm there before you ship.

The current Claude models

	Claude Fable 5.1	Claude Opus 5.5	Claude Sonnet 5.5	Claude Haiku 5.5
Best for	Demanding reasoning, long-horizon agentic work	Long-running agentic coding and knowledge work	Best combination of speed and intelligence	High-volume, latency-sensitive tasks
Claude API ID	claude-fable-5-1	claude-opus-5-5	claude-sonnet-5-5	claude-haiku-5-5
Relative latency	Slower	Moderate	Fast	Fastest
Context window	1M tokens	1M tokens	1M tokens	1M tokens
Max output	128K tokens	128K tokens	128K tokens	128K tokens
Thinking	Adaptive, always on	Adaptive, always on	Adaptive	Adaptive
Default effort (API)	high	medium	high	medium
Input / output price per MTok	$10 / $50	$4 / $20	$2 / $10	From $0.10 / $0.50
Reliable knowledge cutoff	Jun 2026	Jun 2026	Jun 2026	Jun 2026

MTok means one million tokens. Anthropic notes that 1M tokens is roughly 555,000 words on the current tokenizer. All current models accept text and images and produce text, and all support tool use and vision.

Haiku 5.5’s “from” price is tiered by prompt size: $0.10 input and $0.50 output per MTok for prompts up to 100,000 tokens, and $0.50 / $2.50 for prompts over 100,000 tokens. The other current models charge the same rate across the full 1M context.

There is also Claude Mythos 5.1, priced like Fable 5.1, which Anthropic offers only to organisations verified through programs such as its Cyber Verification Program. Most readers can ignore it.

What each model line is for

Claude Haiku: fast and cheap

Haiku is the line for volume. Anthropic positions Haiku 5.5 (released October 7, 2026) for classification, extraction, routing and subagent tasks, where you make many calls and latency matters more than depth. It’s the cheapest current model by a wide margin, and Anthropic’s “efficiency-first” advice is to prototype on Haiku and upgrade only where it falls short.

Claude Sonnet: the everyday workhorse

Sonnet is the middle of the range and the most searched name in the family. Sonnet 5.5 is described as “the best combination of speed and intelligence”, with suggested uses including code generation, data analysis, content creation, visual understanding and agentic tool use. At $2 / $10 per MTok it costs half of Opus 5.5, so it’s the natural candidate for production traffic once you’ve proved quality.

Claude Opus: deep work and agents

Opus is the line for hard, long tasks. Anthropic’s model-selection guide says “Most workloads start with Claude Opus 5.5”, listing multi-hour autonomous coding agents, large refactors, complex systems engineering, vision-heavy workflows and computer use as typical fits. Opus 5.5 is also cheaper per token than earlier Opus models ($4 / $20 versus $5 / $25 for Opus 5 and the 4.x versions).

Claude Fable: the top tier

Fable sits above Opus. Anthropic calls Fable 5.1 its most capable model open to all customers and recommends it for agent sessions that run for hours, multistep deep research and analysis carried through to a finished document. It costs 2.5x Opus 5.5 per token, so Anthropic suggests moving to it only when Opus 5.5 at high effort still falls short on your evals.

How to choose a Claude model

Anthropic’s choosing a model guide offers two starting strategies.

Strategy	Start with	Then	Good for
Capability-first	Opus 5.5	Lower effort or try cheaper models once quality is proven; move up to Fable 5.1 only if evals still fail	Complex reasoning, advanced coding, agentic work
Efficiency-first	Haiku 5.5	Upgrade only for specific capability gaps	Prototypes, tight latency, high-volume or cost-sensitive tasks

A few practical rules we’d add:

• Tune effort before switching models. The effort setting (low, medium, high, xhigh, max) trades depth for cost and speed within one model. Anthropic says this is often a better lever than changing models. Note that Opus 5.5 and Haiku 5.5 default to medium on the API, while Fable 5.1 and Sonnet 5.5 default to high.
• Test on your own tasks. Build a small set of real prompts with known good answers and run them through two models. Public benchmarks rarely match your workload.
• Mix models. A common pattern is an orchestrator on Opus that hands bulk work to Haiku or Sonnet workers, so most tokens are billed at the lower rate.
• Judge cost per finished task. A cheaper model that needs more retries isn’t cheaper. Our Claude API pricing guide has worked cost examples.

Using Claude models in the apps vs the API

In claude.ai, Desktop and Mobile

In the Claude apps you choose a model and effort level from the menu next to the send button, and you can change both mid-conversation. The Help Center lists effort control for Haiku 5.5, Sonnet 5.5, Opus 5.5, Fable 5.1 and several earlier models, and notes that thinking can’t be switched off on Haiku 5.5, Sonnet 5.5, Opus 5.5, Fable 5.1 or Opus 5.

Which models you see depends on your plan. Fable 5 and Fable 5.1 are available on all paid plans: they’re part of the plan on Max and Team premium seats (up to 50% of your weekly limit), and run on pay-as-you-go usage credits on Pro and Team standard seats. Anthropic also announced that Haiku 4.5 leaves the app model selector on October 29, 2026 for Free, Pro and Max, replaced by Haiku 5.5; it stays in Claude Code, Team, Enterprise and the API.

Our Claude pricing guide covers plans and limits, and Claude Pro vs Max helps decide whether more Fable and Opus usage is worth paying for.

In Claude Code

Claude Code lets you switch models with /model and set effort per session. Usage counts against your subscription or API billing depending on how you sign in. Our Claude Code pricing guide explains both, and What is Claude Code? is the place to start if you’re new.

On the Claude API

You pass the model ID in each request. A minimal call looks like this:

curl https://api.anthropic.com/v1/messages \
  -H "content-type: application/json" \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -d '{"model": "claude-sonnet-5-5", "max_tokens": 1024,
       "messages": [{"role": "user", "content": "Hello, Claude"}]}'

From the 4.6 generation on, IDs have no date suffix, and Anthropic says every model ID is a pinned snapshot. You can also query limits and capabilities at runtime with the Models API, which returns max_input_tokens, max_tokens and a capabilities object for each model. New to the API? Follow our Claude API tutorial and set up a key with the Claude API key guide.

Where each model is available

All four current models are listed on the Claude API and on these cloud platforms, with these IDs:

Platform	ID format	Example (Sonnet 5.5)
Claude API	Bare ID	claude-sonnet-5-5
Amazon Bedrock	anthropic. prefix	anthropic.claude-sonnet-5-5
Claude Platform on AWS	Bare ID	claude-sonnet-5-5
Claude Models on Google Cloud	Bare ID	claude-sonnet-5-5
Microsoft Foundry	Bare ID (deployment name)	claude-sonnet-5-5

Bedrock and Google-managed Google Cloud set their own prices and retirement dates; Foundry follows the Claude API schedule. For Anthropic’s own platforms, retirement is committed as “not sooner than” September 2027 for Fable 5.1, Opus 5.5 and Sonnet 5.5, and October 7, 2027 for Haiku 5.5.

Legacy and retired models

Anthropic still serves several earlier models, listed as legacy: Fable 5, Opus 5, Opus 4.8, Opus 4.7, Opus 4.6, Opus 4.5, Sonnet 5, Sonnet 4.6 and Haiku 4.5. Sonnet 4.5 is marked deprecated.

Legacy model	Input / output per MTok
Fable 5	$10 / $50
Opus 5, Opus 4.5 to 4.8	$5 / $25
Sonnet 5	$2 / $10
Sonnet 4.5 and 4.6	$3 / $15
Haiku 4.5	$1 / $5

Opus 4.1, Opus 4, Sonnet 4 and Haiku 3.5 are retired on the Claude API, though some remain on Bedrock or Google Cloud. If you’re on any legacy model, moving to the 5.5 generation is usually both an upgrade and a price cut. Two things to plan for: newer models use a tokenizer that produces roughly 30% more tokens for the same text, and some request settings (such as disabling thinking or forcing a specific tool) are rejected on the newest models. Read Anthropic’s migration guide before switching IDs in production.

Keep learning

• Compare costs in depth: Claude API pricing.
• Understand the wider product family: What is Claude?
• Build agents on these models: Claude agents guide.
• Get better answers from any model: Claude prompting best practices.
• See how Claude stacks up against OpenAI: Claude vs ChatGPT.