Claude 5 is here: Fable 5, Opus 4.8, Sonnet 5 — and what they cost

Anthropic didn’t ship one new model this cycle — it shipped a whole family at once. Four models, one generation, and the question that actually matters if you’re building anything on top of them: what do you get, and what does it cost you per request.

The lineup at a glance

• Claude Fable 5 — the most capable model in the family, built for the hardest reasoning and long-horizon agentic work. 1M token context window, up to 128K tokens of output.
• Claude Opus 4.8 — the top of the Opus tier, highly autonomous, strong on long agentic runs and knowledge work. Also 1M context.
• Claude Sonnet 5 — the speed-to-intelligence sweet spot, now reaching what used to be Opus-tier quality on coding and agentic tasks.
• Claude Haiku 4.5 — the fast, cheap option for simple, high-volume work. Smaller context window at 200K tokens.

Fable 5 is the one worth understanding

It’s Anthropic’s most capable widely released model, and it works differently from the others in a few ways that actually matter if you’re building with it directly. Thinking is always on — there’s no way to turn it off, the model reasons through every request whether you ask it to or not. It doesn’t support the older “prefill” trick some tools use to force a specific output format. And it requires 30-day data retention on the account using it, so it’s not available under a zero-retention setup.

None of that is a downside exactly — it’s a model built for long, autonomous, high-stakes work, and the defaults reflect that. It’s just not a drop-in swap for a lighter model without checking what changes.

What it actually costs to run

Pricing is per million tokens, split between what you send in and what the model sends back:

• Fable 5: $10 in / $50 out
• Opus 4.8: $5 in / $25 out
• Sonnet 5: $3 in / $15 out (introductory pricing of $2 in / $10 out runs through August 31, 2026)
• Haiku 4.5: $1 in / $5 out

For most day-to-day automation work — the kind of thing we cover in the Automation Tutorials pillar — Sonnet 5 is the model that makes sense: near-top-tier coding and agentic performance at a third of Fable 5’s output cost, with an introductory price cut on top through the end of August. Save Fable 5 for the tasks that genuinely need the ceiling — long-horizon, high-complexity, get-it-right-the-first-time work — rather than defaulting to it out of habit.

Worth watching: this kind of full-family refresh is exactly the move competitors in the space are under pressure to match. We’re holding off on naming specific competitor responses here until there’s something concrete to report — not vague industry chatter — so expect a follow-up in this pillar once that happens.