On Monday, Anthropic announced the latest version of its flagship model, Opus 4.5. This is the last model in the Anthropic 4.5 series to be released, following the launch of Sonnet 4.5 in September and Haiku 4.5 in October.
As expected, the new version of Opus delivers state-of-the-art performance across a variety of benchmarks, including coding benchmarks (SWE bench and terminal bench), tool usage (tau2 bench and MCP Atlas), and general problem solving (ARC-AGI 2, GPQA Diamond).
Notably, Opus 4.5 is the first model to score above 80% in SWE-Bench validation, a popular coding benchmark.
Anthropic also launched a number of parallel products to highlight the computer use and spreadsheet capabilities of Opus and demonstrate how the model performs in these settings. Along with Opus 4.5, Anthropic is making its previously pilot Claude for Chrome and Claude for Excel products more widely available. The Chrome extension is available to all Max users, and the Excel-focused model is available to Max, Team, and Enterprise users.
Opus 4.5 also includes memory improvements for long context operations, which required significant changes to how models manage memory.
“Training with Opus 4.5 has improved the quality of common long contexts, but context windows alone are not enough,” Dianne Na Penn, head of research product management at Anthropic, told TechCrunch. “Knowing the right details to remember is very important as a supplement to simply having a longer context window.”
These changes also enable a long-requested “endless chat” feature for paid Claude users, allowing chat to continue uninterrupted when a model reaches the context window. Instead, the model compacts context memory without warning the user.
tech crunch event
san francisco
|
October 13-15, 2026
Many of the upgrades have been done with agent use cases in mind, specifically scenarios where Opus acts as a lead agent directing a group of Haiku-powered subagents. Managing these tasks requires strong working memory commands, and this is where the memory improvements Penn describes have real value.
“Fundamentals like memory are really important here because Claude needs to be able to explore code bases and large documents, and he needs to know when to go back and look at something again,” Penn says.
Opus 4.5 will face stiff competition from other recently released Frontier models, particularly OpenAI’s GPT 5.1 (released November 12) and Google’s Gemini 3 (released November 18).
Source link
