Rate
Claude models explained: Fable 5.1, Opus 5.5, Sonnet 5 and Haiku 4.5
Anthropic ships Claude in four sizes that are available to everyone, plus a restricted tier for vetted organizations. Here is what each one is for and what it costs, as of September 2026.
The lineup at a glance
| Model | Best for | API input / output per 1M tokens | Context window | Released |
|---|---|---|---|---|
| Claude Fable 5.1 | The hardest reasoning and long agentic work | $10 / $50 | 1M tokens | Sep 1, 2026 |
| Claude Opus 5.5 | Most workloads, especially coding and agents | $4 / $20 | 1M tokens | Sep 22, 2026 |
| Claude Sonnet 5 | Speed and intelligence balanced | $2 / $10 | 1M tokens | Jun 30, 2026 |
| Claude Haiku 4.5 | Fast, cheap, real-time tasks | $1 / $5 | 200K tokens | Oct 15, 2025 |
All four offer a 50% discount through the Batch API, and cached input is billed far below the normal input price.
Claude Opus 5.5: the default pick
Opus 5.5 is the newest model and the one Anthropic recommends “for most workloads.” Anthropic says it performs at the level of Fable 5.1 on most work and costs 40% less to run than Opus 5. At $4 input and $20 output per million tokens it is also cheaper than the previous Opus 5 ($5 / $25).
It’s a strong choice for coding agents and long knowledge-work tasks, where its low cache-read price ($0.20 per million tokens) matters because the same context gets re-read many times.
Claude Fable 5.1: the top of the range
Fable 5.1 is Anthropic’s most capable generally available model, aimed at “demanding reasoning and long-horizon agentic work.” It costs $10 / $50 per million tokens, the most expensive in the lineup.
Two details matter for your bill. Like all Claude 4.7-and-later models, Fable 5.1 uses a newer tokenizer that produces about 30% more tokens for the same text than older Claude models, so a prompt costs more than older per-token comparisons suggest. On the other hand, cache reads on Fable 5.1 cost just $0.25 per million tokens, which Anthropic says makes it up to 45% cheaper than Fable 5 for agentic work.
Claude Sonnet 5: the everyday model
Sonnet 5 is the default model on claude.ai’s Free and Pro plans. It launched at an introductory $2 / $10 price, and Anthropic later made that price permanent. For most chat and writing tasks it is the best value in the lineup.
Claude Haiku 4.5: cheap and fast
Haiku 4.5 is the smallest model: $1 / $5 per million tokens and a 200K-token context window. Use it for classification, extraction, routing and anything that needs a quick answer at high volume.
Fable vs Mythos, and Project Glasswing
Fable and Mythos are the same underlying model with different safeguards. Fable has stronger safeguards for general use and is what you can use on claude.ai and the API. Mythos has fewer safeguards and is limited to vetted organizations: Mythos 5.1 is available only through Anthropic’s Cyber Verification and Life Sciences Verification programs, currently for US organizations.
Separately, Claude Mythos Preview, which is not publicly available, was given to partners in Project Glasswing, a defensive cybersecurity effort Anthropic launched on April 7, 2026 with companies including AWS, Apple, Google, Microsoft and NVIDIA. Glasswing partners can now upgrade to Mythos 5.
The June 2026 suspension
On June 12, 2026, after a US government export-control directive covering foreign nationals, Anthropic suspended access to Fable 5 and Mythos 5 for all customers. The controls were lifted on June 30, and Fable 5 became available globally again on July 1. Other Claude models were not affected.
Which Claude plan do you need?
On claude.ai, the Free plan includes Sonnet and Haiku. Pro costs $20 a month ($17 a month billed yearly) and adds Opus and Fable. Max starts at $100 a month for much higher usage limits. See every plan side by side on our plans page.
Quick recommendations
- Building a coding agent? Start with Opus 5.5.
- Hardest problems, budget no object? Fable 5.1.
- A chatbot or writing tool on a budget? Sonnet 5.
- Millions of small tasks? Haiku 4.5.
Estimate your own bill in the cost calculator, and see how people rate each model today on the dumb meter.