On September 22, 2026 (US time), Anthropic announced Claude Opus 5.5 through its official claudeai account. It is the first model in the brand-new Claude 5.5 series. According to Anthropic, Opus 5.5 matches Claude Fable 5.1 on most tasks while running 40% cheaper than Opus 5.
On pricing, AInvest, citing official sources, reports that Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens. Note the two different figures: the list price is about 20% lower than Opus 5's $5/$25, while the official "40% lower running cost" is a separate claim — Anthropic has not explained how it is measured, so the two numbers should not be conflated.
Why a "5.5 series"
When Opus 5 launched on July 24, it followed the playbook of "near-Fable 5 performance at half the price ($5/$25)." Two months later, Anthropic skipped a small Opus 5.1-style iteration and opened the "5.5 series" instead, with its first model aimed directly at Fable 5.1, released September 1 at $10/$50.
A week before the launch, rumors of a model codenamed claude-wafer-eap were already circulating, and traders on the prediction platform Polymarket were betting with 83% odds that a new Opus would ship before September 24. The official announcement settles the speculation.
How the three models differ
| Model | Released | API pricing (input/output) | Official performance claim |
|---|---|---|---|
| Claude Opus 5 | 2026-07-24 | $5/$25 | Near Fable 5 performance |
| Claude Opus 5.5 | 2026-09-22 | $4/$20 (reported) | Fable 5.1-level on most tasks, 40% lower running cost than Opus 5 |
| Claude Fable 5.1 | 2026-09-01 | $10/$50 | Anthropic's most capable model, 75% cheaper cache reads |
For details on Fable 5.1, see our earlier piece Claude Fable 5.1 Released: Long Task Upgrade, Three Changes for Migration (/en/article/1967).
What it means for enterprise model selection
Fable 5.1-level capability is now available at roughly 40% of the price ($4/$20 vs $10/$50). For enterprises, Fable 5.1's role shrinks to long-horizon tasks that genuinely need top-tier capability. Opus 5 is in the most awkward spot — weaker than 5.5, yet priced higher.
This continues Anthropic's 2026 playbook of making cost reduction the headline itself: Opus 5 won share in July with "near-Fable 5 performance at half the price," Fable 5.1 kept long-task users in September with 75% cheaper cache reads, and now Opus 5.5 extends the cost push into the 5.5 series.
One caveat: Anthropic has so far only disclosed the performance claim and the cost-reduction figure. Benchmark scores, the API model ID, and availability channels (API, Claude app, Claude Code) are undisclosed. Don't assume it's available everywhere until the official docs are updated.
FAQ
Q: What's the biggest difference between Opus 5.5 and Opus 5?
A: It's not just more for more money. On Anthropic's numbers, Opus 5.5 moves the performance target from Fable 5 up to Fable 5.1, runs 40% cheaper than Opus 5, and drops API pricing from $5/$25 to $4/$20. For existing Opus 5 users, it's a direct in-series upgrade.
Q: I'm already on Fable 5.1 — should I switch?
A: Fable 5.1 remains Anthropic's most capable model; long-horizon and top-tier workloads stay its territory. But if Anthropic's "Fable 5.1-level on most tasks" claim holds up in independent benchmarks, plenty of everyday workloads can move down from Fable 5.1 to Opus 5.5 and cut costs dramatically. Wait for third-party evaluations before migrating.
Q: Can I use Opus 5.5 now?
A: Anthropic hasn't disclosed the API model ID or the channel list. Until the official docs or announcements are updated, go by the API documentation and don't trust third-party "already live" claims.
