Every Claude model explained, from Haiku 5.5 to Mythos and Fable

Newest model

Claude Haiku 5.5, October 7, 2026

Most capable for everyone

Claude Fable 5.1

Restricted

Claude Mythos 5.1, for approved security and science teams

Plans

Free (Sonnet, Haiku), Pro $20/mo, Max from $100/mo

Until this year Claude came in three sizes: Haiku, Sonnet and Opus. In June 2026 Anthropic added Fable and Mythos above Opus, and both run on the same underlying model. Fable keeps safeguards that hand some requests to Opus 4.8. Mythos has some of those lifted and goes only to approved security and science teams. The current versions, 5.1, came out on September 1.

The current Claude lineup

* For prompts up to 100,000 tokens. Longer prompts cost $0.50 in and $2.50 out.

Mythos vs Fable: same model, different rules

Anthropic introduced the first versions, Claude Fable 5 and Claude Mythos 5, on June 9, 2026, and the 5.1 versions in the table above on September 1. In Anthropic's words, Fable is a "Mythos-class model that we've made safe for general use," while Mythos has "the safeguards lifted in some areas."

Access to both models was suspended on June 12, three days after launch, and restored on July 1, according to a note on Anthropic's launch page.

Unless you're in one of those programs, Fable is the one you can use. The gap shows up in security, biology and chemistry work, where Fable's classifiers step in. Anthropic says most biology and chemistry requests on Fable currently go to Opus 4.8, even though fallback happens in fewer than 5% of sessions overall.

Opus 5.5, Sonnet 5.5 or Haiku 5.5?

These three are the models most people use day to day. Anthropic published these scores for them:

Sonnet 5.5 scores 4.2 points higher than Opus on Terminal-Bench 4.0 and trails it by 2 Elo points on GDPval. Anthropic still says Opus 5.5 "remains clearly stronger at complex, open-ended work requiring sustained judgment," and that it "performs at the level of Claude Fable 5.1 on most work." How we'd pick between them:

• Opus 5.5 for long projects where the model has to make its own calls: planning, research, large refactors. It's also about 40% cheaper to run than Opus 5 on typical workloads and writes output more than 30% faster.
• Sonnet 5.5 for well-defined jobs: fixing a bug, drafting a document, building slides or a spreadsheet. It's 30% faster than Sonnet 5 and up to 30% cheaper per task.
• Haiku 5.5 when you run the same task thousands of times. Anthropic calls it "the cheapest, fastest, and most capable small model we've ever released," and it's the first Haiku with an adjustable effort setting.

Claude pricing: which plan gets which model

Since June 23, Fable on the Pro plan draws on usage credits instead of being included. The Opus 5.5 launch raised the five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans. Max and Team subscribers also get monthly API credits that work with any model: $100 on Max 5x, $200 on Max 20x and up to $500 pooled on Team.

On the ChatGPT side, the plan decides the model too: see which GPT-6 model each ChatGPT plan gets. And Anthropic itself filed to go public on October 26; our Anthropic IPO page covers the S-1 and the expected date.

Claude FAQ