Credit: Anthropic

Anthropic has launched Claude Opus 4.5, calling it its smartest, most efficient and most capable model to date. Opus 4.5 advances software engineering, long-running agents, research workflows and everyday productivity, and is now available across Anthropic’s apps, API and all major cloud providers. Developers can access it using claude-opus-4-5-20251101, priced at $5 / $25 per million tokens.

A leap forward in real-world software engineering

Opus 4.5 posts state-of-the-art results on SWE-bench Verified and leads across seven of eight programming languages on SWE-bench Multilingual. Early testers said the model “gets it,” reasoning cleanly about tradeoffs, navigating ambiguity and solving multi-system bugs with minimal hand-holding. Anthropic noted that Opus 4.5 even outperformed every human candidate on their two-hour internal engineering take-home exam.

Credit: Anthropic

The model also shows stronger performance in reasoning, math, vision and multilingual tasks, and demonstrates creative, policy-aware problem solving expected from agentic systems.

Stronger alignment and resistance to attacks

Opus 4.5 is described as Anthropic’s most robustly aligned model so far. The company highlights improvements in avoiding unintended behavior and resisting advanced prompt injection attacks. According to internal evaluations and Gray Swan’s testing, Opus 4.5 is harder to trick than any other frontier model in its class.

New control and efficiency on the Claude Developer Platform

Developers now get more control through a new effort parameter in the Claude API. Teams can choose between faster responses or deeper, higher-effort reasoning. At medium effort, Opus 4.5 matches Sonnet 4.5’s top results while using 76 percent fewer output tokens. At maximum effort, it surpasses Sonnet 4.5 while still using 48 percent fewer tokens.

Opus 4.5 also benefits from improved context management, memory and tool use, enabling it to run longer tasks with less oversight. In deep research evaluations, these upgrades delivered a nearly 15-point performance boost.

Platform and product updates

Anthropic is rolling out new tools, integrations and app improvements:

• Claude Code now builds clearer, more detailed plans and executes tasks more thoroughly. It is also available in the desktop app for parallel multi-agent workflows.

• Claude apps now support long, uninterrupted conversations thanks to automatic context summarization.

• Claude for Chrome is available to all Max users for cross-tab automation.

• Claude for Excel expands to all Max, Team and Enterprise users.

• Usage limits for Opus 4.5 have increased for Max and Team Premium, replacing previous Sonnet-based allocations.

Anthropic positions Opus 4.5 as a preview of how AI systems will reshape complex work. As the model becomes widely adopted for coding, research and computer use, the company plans further updates across its product suite.