Haiku, Sonnet, or Opus: Which Claude Should You Actually Use
Open claude.ai and there's a model picker sitting right above the message box, and most people never touch it. Anthropic currently ships four Claude tiers at once — Haiku 4.5, Sonnet 5, Opus 4.8, and the separately-billed Mythos-class Fable 5 — and each one trades speed, cost, and depth differently. Here's what actually separates them and when switching is worth it.
Same underlying family, three different trade-offs between speed, depth, and cost.
Same Family, Three Different Jobs
Anthropic names its Claude tiers after poem forms on purpose: Haiku is short, Sonnet is mid-length, Opus is the longest and most elaborate. The naming maps directly onto the models — Haiku 4.5 is the smallest and fastest, built for high-volume, low-latency work; Sonnet 5 is the balanced middle tier and Anthropic's default choice for most people; Opus 4.8 is the largest, reserved for problems where depth of reasoning matters more than speed or price. The number after each name is the generation, not the size — Haiku 4.5, Sonnet 5, and Opus 4.8 are all current, just released at different points.
Above all three sits a newer, separately-branded tier: Mythos-class models, currently represented by Claude Mythos 5 and Claude Fable 5. They share one underlying model, but Fable 5 carries additional safety measures around biology, cybersecurity, and LLM research and development, and is the version generally available; Mythos 5 itself is used in more limited, higher-trust deployments.
What Actually Changes Between Them
Haiku 4.5
The fastest and least expensive tier. Best for high-volume tasks — quick edits, simple classification, routing, and drafts where speed matters more than depth.
Sonnet 5
The everyday driver. Handles the large majority of writing, research, analysis, and coding at a level close to Opus, for meaningfully less cost and latency.
Opus 4.8
The flagship for genuinely hard problems: long agentic coding chains, unfamiliar large codebases, and multi-step technical reasoning where a wrong turn is expensive to unwind.
Fable 5 (Mythos-class)
Billed and accessed separately from the standard three. Sits above Opus for organizations that need Anthropic's most capable model with extra safeguards attached.
Pricing and exact context-window limits shift often enough that repeating specific numbers here would likely be stale within weeks — check claude.com/pricing for the current figures before making a decision that depends on them. What's stable is the relative ordering: Haiku is always cheapest and fastest, Opus is always the most expensive and most capable of the standard three, and Sonnet sits between them by design.
How to Actually Choose
For most day-to-day use on claude.ai — writing, research, coding help, general questions — Sonnet 5 is the right default and the one Anthropic points people toward automatically. The gap between Sonnet and Opus on everyday tasks tends to be small enough that most people won't notice it, which is exactly why Sonnet, not Opus, is the model selected by default when you open a new chat.
Reach for Opus specifically when the task is long, technical, and the cost of a mistake is high: refactoring a large unfamiliar codebase, an agentic workflow that runs many steps without a human checking in between, or reasoning that has to hold up across a genuinely long chain of dependent steps. Reach for Haiku when you're running the same kind of request at scale — batch-processing hundreds of short items, or building a feature where response time matters more than nuance. Claude Code's smart model switching already does a version of this automatically, routing simple file reads and boilerplate to Haiku while reserving Sonnet and Opus for heavier work, which is a reasonable model for how to think about it manually too.
Bottom Line
Start on Sonnet 5 for basically everything. Drop to Haiku 4.5 when you're doing high-volume, low-stakes work and want speed. Move up to Opus 4.8 only when a task is genuinely hard — deep technical reasoning, long agentic chains, or code you can't afford to get subtly wrong. Fable 5 is a separate purchase decision for organizations with a specific need for Anthropic's most capable model, not something an individual user typically reaches for by default.
More AI Guides, Written Clearly
Browse our full library of AI tool guides, updated regularly as models and features change.
Common Questions
What's the difference between Claude Haiku, Sonnet, and Opus?
They're three sizes of the same model family, named after poem forms to signal scale: Haiku is the smallest and fastest, Sonnet is the balanced middle tier, and Opus is the largest and most capable. The number after each name (4.5, 5, 4.8) marks the generation.
Which Claude model should I use for everyday chat?
Sonnet 5. It's the default model on claude.ai for a reason — it handles most writing, research, analysis, and coding tasks at a level very close to Opus, at meaningfully lower cost and higher speed.
When is it worth paying for Opus instead of Sonnet?
For the hardest reasoning and coding problems — long agentic tasks, large unfamiliar codebases, multi-step technical work where a wrong turn is expensive. For everyday writing, summarizing, or straightforward code, the gap over Sonnet is small enough that most people won't notice it.
What is Claude Fable 5 and how is it different from Opus?
Fable 5 sits in Anthropic's new Mythos tier above Opus, billed and accessed separately. It shares its underlying model with Claude Mythos 5, but Fable 5 carries additional safety measures around biology, cybersecurity, and LLM research and development.
More from the creator
Explore two other projects from the same creator.