Anthropic now sells more Claude tiers than most teams can keep straight. There is a flagship for hard work, a restricted version for vetted specialists, a workhorse Opus line, a balanced Sonnet line and a cheap Haiku line.
This page lists them with prices and the job each one suits, checked on 7 November 2026.
For the three-vendor view, see the AI model comparison for 2026. For the release list across all vendors, see latest AI models.
The lineup
| Model | Listed | Context | Price (input / output) | Best for |
|---|---|---|---|---|
| Claude Fable 5.1 | 1 Sep 2026 | 1M | $10 / $50 | Hardest coding, writing, long agents |
| Claude Mythos 5.1 | Sep 2026 | 1M | Restricted | Vetted cyber and life-science work |
| Claude Opus 5.5 | 22 Sep 2026 | 1M | $4 / $20 | General frontier work at lower cost |
| Claude Sonnet 5.5 | 28 Sep 2026 | 1M | $2 / $10 | Balanced everyday work |
| Claude Haiku 5.5 | 7 Oct 2026 | 1M | $0.10 / $0.50 | High-volume, low-complexity tasks |
| Claude Opus 5 | 24 Jul 2026 | 1M | $5 / $25 | Previous Opus |
| Claude Fable 5 | 9 Jun 2026 | 1M | $10 / $50 | Previous Fable |
| Claude Opus 4.8 | 27 May 2026 | 1M | $5 / $25 | Previous Opus |
| Claude Sonnet 4.6 | June 2026 | 1M | $3 / $15 | Previous Sonnet |
| Claude Haiku 4.5 | Oct 2025 | 200K | $1 / $5 | Previous Haiku |
Prices are list prices per million tokens. Batch tiers are generally half price.
Fable 5.1: the flagship
Fable 5.1 is the model Anthropic positions for the hardest coding, knowledge work and long-running tasks. The list price is unchanged from Fable 5 at $10 / $50, but cache reads fell to $0.25 per million tokens.
Anthropic estimates that cuts typical workload costs by around 25%, and up to around 45% for highly agentic work where cache reads dominate. Our Fable 5.1 upgrade notes cover the detail.
It is overkill for routine work and fairly priced for work where a wrong answer is expensive.
Mythos 5.1: the restricted twin
Mythos 5.1 is the same model with fewer safeguards, available only through trusted access programmes for cybersecurity and life-science organisations. It is not a general purchase.
The Mythos 5.1 guide explains who qualifies and what changes.
Opus 5.5: the workhorse
At $4 / $20, Opus 5.5 is half the price of Fable 5.1 and still a frontier-class model. For most production workloads that need strong reasoning but not the absolute ceiling, this is the sensible default.
It is also the tier to test when Fable’s price is hard to justify. On the benchmarks Anthropic published, Opus 5 sits just behind Fable 5.1: 52.3% versus 55.8% on Terminal-Bench 4.0, and 75.4% versus 77.9% on OSWorld 2.0 partial.
Sonnet 5.5: the balanced tier
At $2 / $10, Sonnet 5.5 is the volume workhorse. It suits support automation, content drafting, summarisation and the bulk of agent steps, with escalation to Opus or Fable for hard cases.
This is the tier most product teams should start with.
Haiku 5.5: the cheap tier
At $0.10 / $0.50, Haiku 5.5 is remarkably cheap for a 1M-context model. It fits classification, extraction, routing, tagging and simple replies. Do not use it for judgement-heavy work.
How to choose
| If your work is… | Start with |
|---|---|
| Routine, high volume | Haiku 5.5 |
| Everyday product work | Sonnet 5.5 |
| Strong reasoning at reasonable cost | Opus 5.5 |
| Hardest coding, writing or agents | Fable 5.1 |
| Vetted security or life sciences | Mythos 5.1 |
The practical pattern is a ladder: Haiku for the bulk, Sonnet for normal requests, Opus for hard ones, Fable for the few tasks that justify it. Escalate on confidence, not on habit.
What changed in 2026
- June: Fable 5 and Sonnet 4.6 shipped.
- July: Opus 5 shipped at $5 / $25.
- September: Fable 5.1 and Mythos 5.1 shipped, with cheaper cache reads.
- Late September: Opus 5.5 and Sonnet 5.5 shipped.
- October: Haiku 5.5 shipped at $0.10 / $0.50.
The pattern is faster point releases and steady price pressure, especially at the low end.
FAQ
Which Claude model is best?
Fable 5.1 for the hardest work, Opus 5.5 for a balance of quality and cost, Sonnet 5.5 for everyday product work, Haiku 5.5 for volume.
Is Fable 5.1 worth $10 / $50?
Only when a failed task is expensive. For routine work, Opus 5.5 or Sonnet 5.5 is usually the better buy.
What is the difference between Fable 5.1 and Mythos 5.1?
The same model with different safeguards. Fable is general; Mythos is restricted to vetted cyber and life-science organisations.
Do older Claude models still work?
Yes, but they are usually more expensive per token than the newer tiers. Migrate unless you have a specific compatibility reason.
Next step
If you are choosing a Claude tier for a product, our AI app development cost guide covers the build side. For a short evaluation, contact 24Bit System.
Related reading: Claude Fable 5.1 and GPT-6.1 vs Claude Fable 5.1.
