Which Claude Is Which: Opus 4.8 and the Claude 5 Family
Anthropic ships several Claude models at once, and the names don't line up in a tidy 1-2-3 order. As of July 2026 the flagship is Claude Opus 4.8, while the…
Claude guide · as of July 10, 2026 · 4 minutes read · details change — confirm current specs on claude.ai
Anthropic ships several Claude models at once, and the names don't line up in a tidy 1-2-3 order. As of July 2026 the flagship is Claude Opus 4.8, while the newer Claude 5 family adds Sonnet 5 and Fable 5, and Haiku 4.5 holds down the fast-and-cheap end. This guide is about picking the right one for the job instead of always grabbing the biggest.
The short version: most people should default to a mid-tier model and only escalate to the flagship when a task is genuinely hard. Bigger is not automatically better — it's slower and eats more of your usage limit.
The lineup at a glance
| Model | Where it sits | Best for | Trade-off |
|---|---|---|---|
| Claude Opus 4.8 | Flagship | Hard reasoning, long documents, tricky code, multi-step agent work | Slowest, uses the most of your limit |
| Sonnet 5 | Everyday workhorse | Most daily writing, research, coding, chat | Occasionally needs a nudge on the hardest problems |
| Fable 5 | Newer Claude 5 model | Check the current model page for how Anthropic positions it | Least familiar name — verify before you commit a workflow to it |
| Haiku 4.5 | Fast / cheap tier | Quick answers, high-volume simple tasks, snappy back-and-forth | Not for deep reasoning or long, subtle work |
Don't quote yourself exact prices from memory — check current plans and limits (as of July 2026 — confirm at claude.ai / anthropic.com), because tiers and caps change.
Opus 4.8 — the heavyweight
Opus 4.8 is the model to reach for when the problem is actually hard: dense legal or technical documents, a knotty bug spread across many files, planning that has to hold together over many steps, or anything where a shallow answer would cost you real time to fix. It also pairs well with the very large context window (up to roughly 1M tokens on some models), so it can hold a lot of material at once.
The cost is speed and usage. On a simple question, Opus is overkill — you'll wait longer and burn through more of your allowance for an answer a lighter model would have nailed.
Sonnet 5 — the everyday default
For most people, most of the time, Sonnet 5 is the right starting point. It handles the bulk of real work — drafting, summarizing, everyday coding, research, ordinary conversation — at a noticeably quicker pace than the flagship. A good habit: start here, and escalate to Opus 4.8 only if you can feel the model straining.
Fable 5 — the newer name
Fable 5 is the least familiar model in the current lineup, and it's the one to be most careful about describing. Rather than guess at its specialty, check the current model page (as of July 2026 — confirm at claude.ai / anthropic.com) for how Anthropic positions it and where it's available. If you're building a repeatable workflow, confirm what Fable 5 is tuned for before you wire it in.
Haiku 4.5 — fast and cheap
Haiku 4.5 is built for speed and volume: quick lookups, short rewrites, tight back-and-forth, or anything you'd run many times where a slower model would be a waste. It's not the one for deep analysis or long, subtle documents — but for "just answer the question," it's often the pleasant choice.
How to pick
A rough decision path:
- Start with Sonnet 5.
- If the answer feels shallow or the task is genuinely hard, move up to Opus 4.8.
- If you're doing something simple at high volume, drop to Haiku 4.5.
- For Fable 5, check its current positioning first.
- Match the model to the task, not to your wish for "the best one."
- Re-check the lineup periodically — names and tiers shift.
When the flagship is the wrong tool
Reaching for Opus 4.8 by reflex is a common mistake. Skip it when:
- The task is simple. A definition, a quick rewrite, a short reply — a lighter model is faster and easier on your limit.
- You need something Claude can't do at all. Claude has no native image generator (it reads images and can build visuals with code, but it won't paint one like DALL-E or Imagen), and there's no always-on voice mode like ChatGPT Advanced Voice or Gemini Live. A heavier Claude tier doesn't unlock these.
- Privacy or offline is the priority. Every Claude model is cloud-only. If the data can't leave your machine, a local model is the better call — see the guide below.
- You're rate-limited. If you're bumping into caps, spending them on the heaviest model for light work is the wrong trade.
Rule of thumb: the flagship is for hard problems, not for reassurance. Default to the workhorse; escalate on purpose.
