No signs of slowing down: Anthropic launches Claude Opus 5.5, 30% faster than previous model

Anthropic has launched Claude Opus 5.5, its latest artificial intelligence model, with the company putting greater emphasis on lowering the cost of running advanced AI systems while improving their ability to handle coding, research and other complex tasks.

The launch comes just days after Anthropic CEO Dario Amodei called on the AI industry to slow the pace at which it develops increasingly capable models, arguing that safety measures need time to keep up with rapidly advancing capabilities. “We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain,” Amodei wrote in an essay published earlier this month. Anthropic has described Opus 5.5 as its “first release since we called for pacing the frontier,” saying the model underwent external testing before launch and carries safeguards developed for its most capable systems.

The AI company says Opus 5.5 performs at the level of its more powerful Claude Fable 5.1 on most tasks, but costs 40% less to run than the previous-generation Opus 5 on typical workloads. It also generates output more than 30% faster than Opus 5, according to Anthropic.

Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens, compared with $5 and $25, respectively, for Opus 5. Cache reads, which Anthropic says account for a large share of the cost of coding and agentic workloads, have been cut from $0.50 to $0.20 per million tokens.

“Where Opus 5.5’s advantage is very clear is efficiency. It costs less per token than Opus 5 and uses fewer tokens per task, which nets out to a 40% drop in costs,” Anthropic said.

The company is pitching the model particularly for long-running AI agents and software engineering work. In one test cited by Anthropic, Opus 5.5 audited and fixed a 200,000-line codebase in under three hours, compared with more than 20 hours for Opus 5.

Mario Rodriguez, Chief Product Officer at GitHub, said the model used “among the fewest tokens and steps” in the company’s testing across GitHub Copilot CLI and VS Code. “More than making individual tasks more efficient, it’s making developers’ bigger projects more achievable,” he said.

Anthropic puts focus on AI safety with Opus 5.5

Alongside performance and cost, Anthropic is making safety a key part of the Opus 5.5 release. The company says the model achieved its best scores so far on its automated behavioural audit and is more resistant to prompt-injection attacks than Opus 5.

In a new evaluation testing whether models try to cross containment boundaries, Anthropic said Opus 5.5 attempted to circumvent those boundaries around 85% less often than Opus 5 or Claude Mythos 5.1. The company added that every such attempt by Opus 5.5 was low severity and self-reported.

However, Anthropic acknowledged that evaluating increasingly capable models remains difficult, saying that building tests capable of reliably detecting every failure before deployment “remains an unsolved problem.”

Given the model’s cybersecurity and biology capabilities, Anthropic is also deploying additional safeguards. Most cybersecurity requests covered by these protections will be redirected to the less powerful Opus 4.8, while safeguards for certain biology-related tasks are similar to those used for Fable 5.1.

Claude Opus 5.5 is available through Claude and the Claude Platform, as well as through Amazon Web Services, Google Cloud and Microsoft Foundry. Anthropic said Claude Sonnet 5.5 and Haiku 5.5 will follow in the coming weeks.