Claude Opus 4.5 comes alongside updates to the firm's developer platform and consumer applications. The company says that Claude Opus 4.5 achieves an SWE-bench Verified score of 80.9 percent. In comparison, ChatGPT-5.1-Codex-Max achieved a score of 77.9 percent and Sonnet 4.5 achieved 77.2 percent.

Anthropic has unveiled Claude Opus 4.5, an upgrade from Opus 4.1, with claims that it excels in coding, agentic tasks, and computer use. The new release comes alongside updates to the firm's developer platform and consumer applications. The company says that Claude Opus 4.5 achieves an SWE-bench Verified score of 80.9 percent. In comparison, ChatGPT-5.1-Codex-Max achieved a score of 77.9 percent and Sonnet 4.5 achieved 77.2 percent.

Claude Opus 4.5 new features

The company says that Claude Opus 4.5 outscored all human candidates on an internal engineering examination, employing parallel test-time compute within a two-hour limit. The new version brings improved vision, reasoning, and mathematics abilities as well. New features include support for extended agent operations, multi-agent systems, and tools like context compaction and memory functions.

The new model also incorporates an 'effort' parameter, allowing users to balance performance against latency and cost, and enables self-improving agents that refine their skills over iterations.

Anthropic describes the system as the most robustly aligned frontier model available, with heightened resistance to prompt injection attacks compared to rivals.

Claude Opus 4.5: Pricing and availability

The biggest change is the reduction of prices to help users and business with more accessibility. Claude Opus 4.5 is now priced at $5 per million input tokens and $25 per million output tokens, a two-thirds cut from the previous Opus 4.1 rates.

Caude Opus 4.5 is available through the app, the API, and through major cloud platforms. Eligible Max, Team, and Enterprise subscribers benefit from removed Opus-specific usage caps and elevated overall limits, including 'infinite chats' that summarise conversations to bypass context window constraints.

Claude Opus 4.5 can now also be integrated into Excel, a possible iteration through its recent partnership with Microsoft. There is also a Chrome extension for Max users and an updated desktop app.

Is Claude Opus 4.5 as great as it claims?

ZDNet, in its evaluation, questioned Anthropic's claim of supremacy. Claude Opus 4.5 succeeded in only half, failing on tasks involving a WordPress plugin development and a currency string function rewrite due to issues with file handling, incomplete logic, and validation errors. It performed adequately on bug identification in PHP code and integration of AppleScript with other tools.