Claude Opus 5.5, Sonnet 5.5, Haiku 5.5: How to Choose

Anthropic shipped the full Claude 5.5 family in 16 days: Opus 5.5 on September 22, Sonnet 5.5 on September 28 and Haiku 5.5 on October 7, 2026. The short answer: Sonnet 5.5 is the new everyday default, Opus 5.5 is for long and complex jobs, and Haiku 5.5 is for cheap, high-volume tasks.

Prices at a glance

• Claude Haiku 5.5 (October 7): $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. Best for summaries, classification, sub-agents and live support.
• Claude Sonnet 5.5 (September 28): $2 input and $10 output per million tokens. Best for everyday coding, documents, slides and design polish.
• Claude Opus 5.5 (September 22): $4 input and $20 output per million tokens. Best for long migrations, code audits and research reports.

Haiku 5.5 costs $0.50 input and $2.50 output per million tokens once a prompt goes over 100,000 tokens.

What changed in each model

• Opus 5.5 costs 40% less to run than Opus 5 on typical workloads, according to Anthropic, and writes its output more than 30% faster. Its thinking mode can no longer be switched off.
• Sonnet 5.5 keeps Sonnet 5's price but needs fewer tokens per task, which Anthropic puts at up to 30% cheaper per job. It is also more than 30% faster. Since October 7, its cache reads cost $0.10 per million tokens instead of $0.20.
• Haiku 5.5 is the first Haiku with an adjustable effort setting, so you can trade cost against quality. Anthropic says it runs about 75% cheaper than Haiku 4.5 on average.

How to use the Claude 5.5 models

1. In the Claude apps, pick the model in the model menu. The apps default to Medium effort, which is enough for most writing.
2. Through the API, use the model IDs claude-opus-5-5, claude-sonnet-5-5 or claude-haiku-5-5. The API defaults to High effort; lower it for routine work to cut cost.
3. Mix them. Let Sonnet or Opus plan a task and hand repetitive steps, such as pulling one figure out of a report, to Haiku as a sub-agent. Anthropic recommends this pairing for coding work.
4. On WorkCrafter, the text tool and the code tool run Claude next to other engines, so you can compare answers to the same prompt without an API key.

Our tips for Claude 5.5

• Start every new project on Sonnet 5.5 and move up to Opus 5.5 only when a task fails or runs for hours.
• Use Haiku 5.5 for anything you repeat thousands of times: tagging, short summaries, routing.
• Lower the effort level before switching to a cheaper model; it is often the quicker saving.
• If you pay for Claude Max or Team, check the monthly API credit Anthropic announced on October 7: $100 for Max 5x, $200 for Max 20x and up to $500 for Team.

More on Claude 5.5 and WorkCrafter

• The AI Text Tool: A Complete Guide to AI Writing
• Claude vs GPT vs Gemini vs Mistral vs DeepSeek (2026)
• How Much Does AI Generation Actually Cost in 2026?
• See WorkCrafter pricing

Sources on Claude 5.5

• Introducing Claude Opus 5.5 — Anthropic, September 22, 2026. Release date, prices, cost and speed figures, model ID.
• Introducing Claude Sonnet 5.5 — Anthropic, September 28, 2026. Release date, prices, speed and cost figures, default effort levels.
• Introducing Claude Haiku 5.5 — Anthropic, October 7, 2026. Release date, prices, effort setting, Sonnet cache price cut, Max and Team credits.