Anthropic has announced Claude Opus 5.5. According to Anthropic’s announcement on 22 September 2026, Opus 5.5 is positioned as the first release in the Claude 5.5 family, performing at the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5. Anthropic says Opus 5.5 was evaluated by external testers, scored best on its automated behavioural audit, and launches with the company’s most recent safeguards.
Key Details
Performance and benchmarking: Anthropic reports that Opus 5.5 delivers a major step up from Opus 5 and performs comparably to Claude Fable 5.1 on most tasks. In Anthropic’s internal and external benchmarks, Opus 5.5 led in agentic coding, computer use and knowledge work on a range of tests cited by the company.
Cost and speed: Anthropic states Opus 5.5 requires less compute to serve than Opus 5 and will cost roughly 40% less on typical workloads. Published list prices per 1M tokens are: cache reads $0.20 (vs $0.50 for Opus 5), input tokens $4 (vs $5), output tokens $20 (vs $25) and cache writes $5 (vs $6.25). Anthropic also says Opus 5.5 generates output more than 30% faster than Opus 5 and that a fast mode is available at different pricing.
Safety and safeguards: Anthropic reports Opus 5.5 scored best to date on its automated behavioural audit and that it is less likely than recent models to take hard-to-reverse actions or act outside given boundaries. Because Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity in Anthropic’s tests, the company is deploying Opus 5.5 with safeguards similar to those used for Fable 5.1 and will route many high-risk cybersecurity tasks to lower-permission models.
Verification programmes and access: Anthropic says vetted organisations can apply to its Life Sciences Verification Program to use Opus 5.5 for biology research, and the company will expand access to its Cyber Verification Program in the coming weeks for verified cybersecurity practitioners.
Availability and platform support: Anthropic states Opus 5.5 is now available on all platforms, including Amazon Web Services, Google Cloud and Microsoft Azure, and on the Claude Platform. The company also said subscription plans will see increased five-hour usage limits for Pro, Max, Team and seat-based Enterprise plans and that it is providing a rate limit reset for subscription users.
Who May Be Affected
Anthropic’s announcement highlights several groups that may see direct effects: developers and engineering teams using agentic coding tools, cybersecurity practitioners subject to the Cyber Verification Program, life‑science researchers seeking verified access, and enterprises that deploy autonomous agents or model-driven automation. Anthropic cites examples from early testers — such as large codebase migrations and auditing tasks — to illustrate potential efficiency gains.
Why It Matters For Voice AI
Anthropic’s release emphasises two features that organisations building voice or conversational products should note: the company says Opus 5.5 “communicates more naturally” and that early testers found its writing clearer and easier to follow. Anthropic also reports faster output generation and lower per‑token costs, which the company presents as improvements that could affect cost and latency when integrating large models into product workflows.
AiDial analysis: For Australian businesses using voice AI, lower inference costs and faster response generation can reduce running costs for call flows and improve real‑time interactions. Improved clarity and consistency in model output can also reduce the human time needed to review or post‑process responses. These points are our analysis and not claims from Anthropic’s announcement.
What To Watch Next
Anthropic says Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks with similar improvements. The company also notes it will expand the Cyber Verification Program in the coming weeks and that organisations can apply today to the Life Sciences Verification Program. Anthropic references its Opus 5.5 System Card for full evaluation details and points to external pre‑release testing by Frontier Design and METR.
Other operational notes from Anthropic’s announcement include preserved thinking (the anti‑distillation safeguard) applying to API accounts created on or after 31 August 2026, and that Opus 5.5 is available with zero data retention and watermarking measures discussed for EU AI Act compliance.
Sources
Anthropic — Introducing Claude Opus 5.5 (primary source, announcement page and system card links embedded there). Additional Anthropic links cited in the announcement include the Opus 5.5 System Card, Life Sciences Verification Program, Cyber Verification Program and related safety reports.
All facts in this explainer are sourced from Anthropic’s announcement and the linked materials on the Opus 5.5 announcement page.
AiDial analysis: Anthropic frames Opus 5.5 as a performance, cost and safety upgrade relative to Opus 5; Australian businesses that run AI‑driven services should assess whether the model’s pricing, safeguards and verification programmes fit their compliance and operational needs before integration.
This article is general information and not legal advice.
