---
title: Anthropic launches Claude Opus 5.5 with 40 percent lower cost than Opus 5
url: https://www.dataloco.com/en/anthropic-launches-claude-opus-55-with-40-percent-lower-cost-than-opus-5
published: 2026-09-25T16:22:02+00:00
language: en
section: AI
source: https://japan.cnet.com/article/35252851/
organizations: amazon, anthropic, azure, google, microsoft
publisher: Dataloco
---

# Anthropic launches Claude Opus 5.5 with 40 percent lower cost than Opus 5

Anthropic announced Claude Opus 5.5 on September 22 2026. The model matches Claude Fable 5.1 performance while costing 40 percent less than Opus 5. It will be available through Claude Platform and on Amazon Web Services Google Cloud Microsoft Azure. The model identifier is claude-opus-5-5. Opus 5.5 also introduces Claude Sonnet 5.5 and Claude Haiku 5.5 arriving soon. API pricing is 4 dollars per million input tokens and 20 dollars per million output tokens, a 20 percent reduction from Opus 5. Caching reads cost 0.20 dollars per million tokens, 60 percent cheaper than Opus 5. Standard usage costs 40 percent less than Opus 5. Fable 5.1 charges 10 dollars for input and 50 dollars for output, so Opus 5.5 offers comparable performance at less than half the price. Output speed improves by more than 30 percent over Opus 5, and Claude Code and Claude Platform support a fast mode up to 2.5 times faster.

Subscription plans Pro Max Team and Enterprise raise hourly usage limits and provide resetable caps. Benchmarks show Opus 5.5 scoring 66.4 percent on Terminal-Bench 4.0, surpassing GPT-6 Astra at 57.9 percent, Fable 5.1 at 55.8 percent and Opus 5 at 52.3 percent. It also leads GDPval-AA v2.1 with 1846 points above Fable 5.1's 1735. However Opus 5.5 does not top Terminal-Bench-Science 0.1 where GPT-6 Astra scores 64.6 percent. Anthropic acknowledges the performance gap is smaller than benchmark scores suggest. A tester migrated 680,000 lines of code in under a day using Opus 5.5, a task that would take a team weeks. Clio engineers ran an 18 hour continuous task across six repositories, and in internal validation Opus 5.5 succeeded in 39 of 40 attempts to shorten web page load times.

The model places key information first, produces shorter more organized answers and generates more informative slides. Safety tests show Opus 5.5 avoids constraint violations about 85 percent less often than Opus 5 and matches Fable 5.1 in prompt injection resistance. External evaluator METR also tested the model before release. Biological and cybersecurity capabilities reach levels comparable to Claude Mythos 5.1, so Anthropic applies similar safeguards as for Fable 5.1. Regular development tasks can use Opus 5.5 but most cybersecurity work still routes to Opus 4.8. A Cyber Verification Program will soon include Opus 5.5 for expert reviewers.

## This story in other languages

- [Français](https://www.dataloco.com/fr/anthropic-lance-claude-opus-55-avec-performance-equivalente-a-fable-51-a-40-de-prix-reduit)