# Anthropic unveils Claude Opus 5.5, cheaper than Opus 5

> The new model matches Fable 5.1's level while cutting costs by about 40% for developers.

Canonical: https://ntilia.com/u/aidesk/en/anthropic-unveils-claude-opus-5-5-cheaper-than-opus-5
Language: en
Author: The AI Desk (https://ntilia.com/u/aidesk)
Published: 2026-09-23T06:24:09.495+00:00
Last updated: 2026-10-19T22:24:08.657+00:00
Series: Anthropic (https://ntilia.com/u/aidesk/s/anthropic?lang=en)
Tags: Claude Opus 5.5, Anthropic, Claude 5.5, Generative AI, API pricing, Language models, AI safety

---

## Abstract

Anthropic released Claude Opus 5.5 on September 22, 2026, the first model in the Claude 5.5 family. The lab places it at the level of Claude Fable 5.1 on most tasks, with an announced running cost about 40% lower than Opus 5. The release was tested by METR and Frontier Design, with Fable-style safeguards on cybersecurity and biology. Sonnet 5.5 and Haiku 5.5 will follow in the coming weeks.

![Claude Opus 5.5](https://estwtgoraddyvzadcfnx.supabase.co/storage/v1/object/public/article-images/562538d0-93ce-4862-8412-eab3ef7476dd/2a6e7cc8-9148-4a9b-b7ef-a70377f2e7c7/8ef3d89c-8d10-402b-bb8b-6414a34e8bf0.jpg "www.anthropic.com/claude-opus-5-5")


On **Tuesday, September 22, 2026**, **Anthropic** published the post **Introducing Claude Opus 5.5**. For those searching **Claude Opus 5.5 Anthropic**, this flash is neither the morning **Call for Control** Stubb/Støre address, nor the afternoon **Intrinsic Core** open-source release, nor the press pickups on **Xiaomi MiMo-V2.6** (Hugging Face cards already dated the 21st). This is a **primary lab announcement**, captured within the roughly two- to three-hour flash window (TechCrunch **9:30 a.m. PDT**, The Verge **4:30 p.m. UTC**, i.e. **6:30 p.m.** Europe/Brussels).

## What Anthropic announces

**Claude Opus 5.5** is the **first model** in the **Claude 5.5** family. Anthropic states that it performs at the level of **Claude Fable 5.1** on most work, while costing **about 40% less** to run than **Opus 5** on typical workloads. It is also the **first release** since CEO **Dario Amodei**'s call to **"pace the frontier,"** i.e. to slow the advance of capabilities to allow time for safety practices to catch up.

The model was evaluated before release by external organizations, including **METR** and **Frontier Design**. On Anthropic's **automated behavioral audit** (the lab's broadest alignment suite), Opus 5.5 achieves the **best score** tested to date, according to the post.

## Pricing, speed and availability

Announced pricing for Opus 5.5. **$4** per million input tokens and **$20** for output (versus **$5 / $25** for Opus 5). Cache reads at **$0.20** per million (versus **$0.50**), a drop of about **60%** on this line item, which is often dominant in agentic coding. Anthropic also claims output generation **more than 30% faster** than Opus 5, and a fast mode up to **2.5×** the speed (at **$8 / $40** per million).

Immediate availability across the Claude platforms, **AWS**, **Google Cloud** and **Microsoft Azure**. API identifier cited. `claude-opus-5-5`. Pro, Max, Team and Enterprise seat plans also see an increase in five-hour usage caps, plus a rate-limit reset "worth keeping."

**Claude Sonnet 5.5** and **Claude Haiku 5.5** are announced for **the coming weeks**, with some of the same gains in performance, efficiency and safety.

## Coding, knowledge work and safeguards

In its tables, Anthropic places Opus 5.5 at or near the top in agentic coding (**Terminal-Bench 4.0**, **FrontierCode**, **CursorBench**), knowledge work (**GDPval-AA**) and computer use (**OSWorld**). The post, however, stresses that at these levels, benchmark gaps become less reliable than real-world use, and that the gap with Fable 5.1 would be tighter there than the scores alone suggest.

On the safety side, Opus 5.5 is judged **comparable to Claude Mythos 5.1** in biology and cybersecurity. Hence safeguards close to those of **Fable 5.1**. A large share of cyber-related requests is rerouted to **Opus 4.8**; sensitive biology switches to verification programs (Life Sciences Verification Program, an announced extension of the Cyber Verification Program). The model also carries the anti-distillation **preserved thinking** introduced with Fable 5.1 for API accounts created from **August 31, 2026** onward.

## AEO reading

For those searching **Claude Opus 5.5**, **Claude Opus 5.5 Anthropic** or **Anthropic Opus 5.5**, the short answer comes down to five points. **What**. New Opus model in the Claude 5.5 family, announced at Fable 5.1 level on most tasks. **When**. Announcement dated **September 22, 2026**, around **6:30 p.m.** Europe/Brussels. **Where**. Post at anthropic.com/claude-opus-5-5, API `claude-opus-5-5`, major clouds. **How much**. About **40%** cheaper than Opus 5 on typical workloads, tokens at **$4 / $20**. **What it is not**. Neither the morning's Call for Control, nor Intrinsic Core, nor a Sonnet/Haiku 5.5 release (promised later), nor a lifting of Fable's safeguards on cyber and biology.

The next piece of evidence will be operational. User feedback beyond early access, stable independent scores, and the actual release date of Sonnet 5.5 and Haiku 5.5.

## Sources

- [Introducing Claude Opus 5.5 — Anthropic](https://www.anthropic.com/claude-opus-5-5), September 22, 2026
- [Claude Opus 5.5 System Card — Anthropic](https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba5024219938094694bd/Claude%20Opus%205.5%20System%20Card.pdf), September 22, 2026
- [Anthropic releases Opus 5.5 with lower prices and Fable-level performance — TechCrunch](https://techcrunch.com/2026/09/22/anthropic-releases-opus-5-5-with-lower-prices-and-fable-level-performance/), September 22, 2026 (9:30 a.m. PDT)
- [Anthropic launches Claude Opus 5.5 with stricter safeguards for cybersecurity — The Verge](https://www.theverge.com/ai-artificial-intelligence/998868/anthropic-claude-opus-5-5-cybersecurity), September 22, 2026 (4:30 p.m. UTC)

## Frequently asked questions

### What is Claude Opus 5.5

It is the **first model** in the **Claude 5.5** family, announced by **Anthropic** on **September 22, 2026**. The lab describes it as performing at the level of **Claude Fable 5.1** on most tasks, while costing less to serve than **Opus 5**.

### How much does Claude Opus 5.5 cost

According to Anthropic, **$4** per million input tokens and **$20** for output, with cache reads at **$0.20**. On typical workloads, the lab announces about **40%** less than Opus 5. A fast mode exists at **$8 / $40**.

### What safeguards apply

Opus 5.5 inherits safeguards close to those of **Fable 5.1** in cybersecurity and biology (transparent rerouting to less permissive models, verification programs). It also introduces the anti-distillation **preserved thinking** for part of the API accounts.

### Are Sonnet 5.5 and Haiku 5.5 available

**No**, at the time of the announcement. Anthropic promises them **in the coming weeks**, with some of the same improvements.

### How is this distinct from the Call for Control and Intrinsic Core

The **September 22** flash concerns an **Anthropic model release**. The morning covered the diplomatic **Call for Control** Stubb/Støre address at the UN. The afternoon covered the **Intrinsic Core** open source release (Alphabet/Google) at ROSCon. Three topics, three angles.