Anthropic Unveils Opus 4.5: Improved Performance and Longer Conversations for AI Models

This article was generated by AI and cites original sources.

Anthropic has launched its latest Opus 4.5 model, featuring significant enhancements in coding performance and user experience, positioning it as a strong competitor to recent models from OpenAI and Google.

One notable upgrade in the Opus 4.5 release is the extension of conversation durations in consumer applications. This improvement addresses a common critique of the Claude AI model by reducing instances of conversations abruptly ending due to exceeding time limits.

Previously, users faced interruptions when conversations reached a hard context window of 200,000 tokens. Now, Claude employs a summarization process to retain essential information from earlier parts of the chat while eliminating extraneous details, ensuring a more coherent dialogue. Developers utilizing Anthropic’s API can implement similar context management and compaction strategies to enhance user experiences.

Opus 4.5 sets a new benchmark by achieving an accuracy score of 80.9% on the SWE-Bench Verified benchmark, slightly surpassing competitors like OpenAI’s GPT-5.1-Codex-Max and Google’s Gemini 3 Pro. While excelling in agentic coding and tool use assessments, Opus 4.5 slightly trails GPT-5.1 in visual reasoning.

This release marks a significant advancement in AI model capabilities, offering increased performance and more engaging user interactions.

Source: Ars Technica