Haiku to Fable.
Four tiers, one rule: use the smallest model that does the job well, and step up only when the work needs more thinking.
Price per million tokens.
API list prices. A million tokens is roughly 750,000 words. On a Claude plan you pay a flat seat instead.
The trade-off, on one chart.
Position shows each model's place in the lineup as Anthropic describes it, not a benchmark score.
The specs.
| Model | Released | Context | Max output | Speed | Best for |
|---|---|---|---|---|---|
| Claude Haiku 5.5claude-haiku-5-5 | Oct 7, 2026 | 1M tokens | 128K tokens | Fastest | High-volume work: classification, extraction, routing and subagents. From $0.10 / $0.50; prompts over 100K tokens cost more. |
| Claude Sonnet 5.5claude-sonnet-5-5 | Sep 28, 2026 | 1M tokens | 128K tokens | Fast | The best mix of speed and intelligence for everyday tasks, documents and bug fixes. |
| Claude Opus 5.5claude-opus-5-5 | Sep 22, 2026 | 1M tokens | 128K tokens | Moderate | Long-running agentic coding and knowledge work. Anthropic's recommended starting model. |
| Claude Fable 5.1claude-fable-5-1 | Sep 1, 2026 | 1M tokens | 128K tokens | Slower | The hardest reasoning and long-horizon agent work, when Opus 5.5 isn't enough. |
| Claude Mythos 5.1claude-mythos-5-1 | Sep 1, 2026 | 1M tokens | 128K tokens | Slower | Fable-class capability for verified cybersecurity and life sciences organizations. |
Which model for which job.
| Job | Haiku | Sonnet | Opus | Fable |
|---|---|---|---|---|
| Sorting, tagging and pulling data out of documents | ||||
| Customer support replies and routing | ||||
| Writing emails, posts and first drafts | ||||
| Summarizing long reports and meetings | ||||
| Spreadsheets, analysis and financial models | ||||
| Everyday coding and bug fixes | ||||
| Long-running agents and multi-step projects | ||||
| Strategy, research and the hardest problems |
A coach can set up the right model for each workflow in one session.
Prices, context windows and dates are from Anthropic's model and pricing pages, checked 2026-10-18. Haiku pricing is tiered: prompts over 100K tokens cost more. Sources: Claude Haiku 5.5Claude Sonnet 5.5Claude Opus 5.5Claude Fable 5.1
Quick answers.
Which Claude model should I use?
Start with Claude Opus 5.5, which Anthropic recommends as the default for most work. Use Sonnet 5.5 for faster everyday tasks at lower cost, Haiku 5.5 for high-volume jobs like sorting and extraction, and Fable 5.1 only for the hardest reasoning.
What is the difference between Haiku, Sonnet and Opus?
They trade speed and price against depth. Haiku 5.5 is the fastest and cheapest, Sonnet 5.5 balances speed and intelligence, and Opus 5.5 handles long-running agentic and knowledge work. All three have a 1M-token context window.
How much do Claude models cost?
API list prices per million input and output tokens are $0.10 and $0.50 for Haiku 5.5 (prompts up to 100K tokens), $2 and $10 for Sonnet 5.5, $4 and $20 for Opus 5.5, and $10 and $50 for Fable 5.1. Claude plans charge a flat seat instead.
What is Claude Fable?
Claude Fable 5.1, released September 1, 2026, is Anthropic's most capable generally available model, aimed at demanding reasoning and long-horizon agent work when Opus 5.5 isn't enough. It is slower and costs more than Opus.
What is Claude Mythos?
Claude Mythos 5.1 is the same underlying model as Fable 5.1, offered only to verified cybersecurity and life sciences organizations through Anthropic's verification programs. It is not available on regular Claude plans.