Anthropic has introduced Claude Opus 5.5, its latest flagship AI model, with major improvements in performance, safety, efficiency, pricing, and communication compared with Claude Opus 5.
It manages to beat Claude Fable 5.1 in most benchmarks while being much cheaper and faster.
The company says Opus 5.5 is its strongest-performing model to date on its automated behavioral audit. The model was also tested by external evaluators, including Frontier Design and METR, before its release.
Major Performance Improvements
Claude Opus 5.5 represents a significant performance upgrade over Opus 5 and is designed to handle more complex tasks.
Anthropic says early testers reported substantial improvements on demanding workloads. In one test, Opus 5.5 completed a 680,000-line code migration in less than a day. The company said the same work could have taken an engineering team weeks.
The model has also improved at identifying and fixing software inefficiencies. In one test, Anthropic asked Opus 5.5 to reduce loading times across every page of a web application. It succeeded in 39 out of 40 attempts.
By comparison, Opus 5 made smaller improvements in the same test while also changing the application’s behavior.
Opus 5.5 also showed stronger results in game development. In a test where several Claude models were asked to build a game from a single prompt, Opus 5.5 achieved the highest score for graphics quality and overall polish.
Lower Price and Faster Responses
Opus 5.5 is also more efficient to operate than Opus 5, allowing Anthropic to reduce its pricing.
Anthropic’s tests show that Opus 5.5 costs about 40% less than Opus 5 on typical workloads when using default settings.
The new model costs $4 per million input tokens and $20 per million output tokens, representing a 20% reduction compared with Opus 5.
Cache reads are also significantly cheaper. They cost $0.20 per million tokens, which is 60% less than Opus 5. Anthropic says cache reads account for most of the costs associated with agentic and coding workloads.
Opus 5.5 also generates output more than 30% faster than Opus 5.
Stronger Safety and Resistance to Attacks
Safety is another major area of improvement in Opus 5.5.
Anthropic says the model achieved the highest scores to date on its automated behavioral audit, which tests Claude across thousands of simulated scenarios.
Opus 5.5 is less likely than recent models to take actions that are difficult to reverse or operate outside the boundaries set for it. It is also more resistant to prompt injection attacks than Opus 5.
Anthropic has expanded its alignment testing for the new model to include longer tasks, impossible tasks, and scenarios based on real-world incidents.
The company notes that these safeguards still have limitations and has provided further details in the Opus 5.5 System Card.
New Safeguards for Biology and Cybersecurity
Anthropic says Opus 5.5 has capabilities in biology and cybersecurity comparable to Claude Mythos 5.1.
Because of this, the company is deploying it with safeguards similar to those used for Claude Fable 5.1.
Vetted organizations can apply for Anthropic’s Life Sciences Verification Program to use Opus 5.5 for biology research.
Anthropic also plans to expand access through its Cyber Verification Program in the coming weeks. Verified cybersecurity professionals will be able to use the model for cybersecurity work.
Sonnet 5.5 and Haiku 5.5 Coming Soon
Anthropic says Claude Sonnet 5.5 and Claude Haiku 5.5 will launch in the coming weeks.
The company expects both models to bring many of the same improvements introduced with Opus 5.5, including gains in performance, efficiency and safety.
The post Claude Opus 5.5 Beats Fable 5.1 at Much Lower Cost and Faster Speed appeared first on ProPakistani.
