Open claude.ai right now and the model picker shows four names: Sonnet 5, Opus 5, Fable 5, and Haiku 4.5. They're not four versions of the same thing. Each one is a different size of model, built for a different kind of work, and priced and gated differently depending on your plan.

Here's what each one does, when Anthropic released it, and which one to have open depending on what you're working on.

The quick answer

For most day-to-day writing, research, and coding, Sonnet 5 already covers it. It's the default model on Claude's Free and Pro plans. When a task is harder than Sonnet can handle cleanly, switch to Opus 5, the strongest model Anthropic makes generally available and the default on the Max plan. Fable 5 is Anthropic's single most capable model, priced higher and metered separately on subscription plans. Haiku 4.5 is the fastest and cheapest of the four, built for quick, high-volume tasks rather than depth.

Claude Sonnet 5: the one you'll use most

Sonnet 5 launched June 30, 2026, replacing Sonnet 4.6 as Claude's default model on Free and Pro plans. Anthropic built it to be more agentic than earlier Sonnet models: it can plan multi-step work, use tools like browsers and terminals, and stay on task without constant check-ins.

On Anthropic's own benchmarks, Sonnet 5's performance lands close to Opus 4.8 (the previous Opus generation) at a noticeably lower cost, which is why it works as the default rather than a stripped-down option. For emails, drafts, research summaries, everyday coding, and most of what a small business owner or creator would ask Claude to do, Sonnet 5 is the right starting point.

Claude Opus 5: when you need more

Opus 5 launched July 24, 2026, days before this was written, and it's the new default model on Claude Max and the strongest model available on Pro. Anthropic describes it as coming close to Fable 5's intelligence at about half the price, and it leads its own benchmarks on coding and knowledge-work evaluations like Frontier-Bench and GDPval-AA.

Reach for Opus 5 when Sonnet 5 gives you something that's close but not quite right: a coding task with a lot of moving parts, a long document you need analyzed carefully, a piece of research where you want the model to check its own work before handing it back.

Claude Fable 5: the most capable model you can get

Fable 5 launched June 9, 2026. Anthropic calls it a "Mythos-class" model, meaning it shares its underlying capability with Claude's most powerful research model but ships with extra safeguards that route sensitive requests (mainly cybersecurity and biology topics) to Opus 4.8 instead. Anthropic says it's state of the art on nearly all of its internal benchmarks, and in early testing, Stripe reported using it to complete a 50-million-line codebase migration in a day that would otherwise have taken a team over two months.

Fable 5 is priced higher than the other three models, and on subscription plans it's metered through usage credits rather than included at the same level as Sonnet or Opus. It's the model to reach for when the task is genuinely at the edge of what you'd trust an AI to handle unsupervised, not for routine daily work.

Claude Haiku 4.5: fast and cheap for simple work

Haiku 4.5 is Claude's fastest model, matching the older Sonnet 4's performance on coding and agent tasks while running several times faster. Anthropic reports a 73.3% score on SWE-bench Verified, a real-world coding benchmark, which is high for a model built primarily for speed. If you're firing off a lot of small requests, want fast turnaround on simple tasks, or you're building something (a chatbot, an automation) that needs to stay responsive, Haiku is the one built for that.

What about Mythos 5?

You won't see Mythos 5 in your model picker. It's the unrestricted version of Fable 5, reserved for a small group of vetted cybersecurity defenders and biology researchers through Project Glasswing, a partnership Anthropic runs with organizations like AWS, Microsoft, and the US government. Nothing here changes what everyday Claude users have access to.

Free, Pro, and Max: what changes

The plan you're on decides how much of this you get access to, not just which model loads by default.

  • Free: Sonnet 5 by default, no cost.
  • Pro ($20/month): Sonnet 5 by default, with Opus 5 available as the strongest option and limited Fable 5 access.
  • Max ($100 or $200/month): Opus 5 by default, higher usage limits, and more room to use Fable 5.
  • Team and Enterprise: seat-based pricing with admin controls, aimed at organizations rather than individuals.

Model pricing and plan limits shift often enough that it's worth checking Anthropic's current pricing page before you commit to a tier.

Bottom line

Start with Sonnet 5. It's the default for a reason, and it handles most writing, research, and coding tasks well. Move up to Opus 5 when a task needs more careful reasoning than Sonnet is giving you. Save Fable 5 for the handful of tasks where you want Anthropic's most capable model and don't mind paying for it. And if you're building something that needs speed at scale rather than depth, that's Haiku 4.5's lane.

Most people reading this will only ever need to know the first two. That's fine. Claude's model lineup keeps changing, but the habit of picking a model that matches your task is one that carries over no matter what Anthropic calls the next one.

Sources:
anthropic.com/news/claude-sonnet-5 · anthropic.com/news/claude-opus-5 · anthropic.com/news/claude-fable-5-mythos-5 · anthropic.com/claude/haiku · anthropic.com/claude/mythos · claude.com/pricing