Anthropic, a leading AI company, has introduced Opus 4.5, the newest iteration of its flagship model, showcasing improved integrations with Chrome and Excel.
Opus 4.5 marks the final release in Anthropic’s series of 4.5 models, following the launch of Sonnet 4.5 and Haiku 4.5 in preceding months. The model excels in various benchmarks, including exceptional performance in coding metrics like SWE-Bench and Terminal-bench, tool usage benchmarks such as tau2-bench and MCP Atlas, and problem-solving assessments like ARC-AGI 2 and GPQA Diamond. Notably, Opus 4.5 has surpassed an 80% score on the SWE-Bench coding benchmark, affirming its capabilities in software development tasks.
In addition to performance enhancements, Anthropic has highlighted Opus’s improved computer usage and spreadsheet capabilities. The company has also unveiled Claude for Chrome and Claude for Excel, complementary products that align with Opus 4.5’s features and are now more widely accessible to users, with the Chrome extension available for Mac users and the Excel-focused model for Mac, Team, and Enterprise users.
Furthermore, Opus 4.5 introduces memory improvements tailored for long-context operations, necessitating significant memory management adjustments to enhance performance. Dianne Na Penn, Anthropic’s Head of Product Management for Research, emphasized the importance of remembering pertinent details alongside extended context windows for optimal training outcomes.
One notable enhancement is the introduction of the ‘endless chat’ feature for paid Claude users, enabling uninterrupted conversations by compressing context memory without user intervention.
Source: TechCrunch