Anthropic unveils Opus 4.5 with major performance and memory upgrades

Anthropic has introduced Opus 4.5, the latest and final release in its 4.5 series of AI models. The rollout follows earlier launches of Sonnet 4.5 in September and Haiku 4.5 in October.

The company says Opus 4.5 delivers state-of-the-art results across several benchmark categories, including coding tasks (SWE-Bench, Terminal-bench), tool-use evaluations (tau2-bench, MCP Atlas), and general reasoning tests such as ARC-AGI 2 and GPQA Diamond. It is also the first model to score above 80% on SWE-Bench Verified, a widely cited benchmark for code generation accuracy.

Anthropic is pairing the model with upgrades to its productivity tools. With the release of Opus 4.5, the company is expanding access to Claude for Chrome and Claude for Excel. The Chrome extension will now be available to all Max users, while the Excel-focused tool will be offered to Max, Team, and Enterprise subscribers.

Read: Sindh launches Pakistan’s first-ever million-rupee esports championship

The model includes significant improvements in long-context performance, requiring changes to how it handles and manages memory. Speaking to TechCrunch, Anthropic’s head of product management for research, Dianne Na Penn, said long context alone is not enough for high-quality output. “Knowing the right details to remember is really important in addition to just having a longer context window,” she said.

Those changes have enabled a long-requested “endless chat” feature for paying Claude users. Instead of stopping when the model reaches its context limit, Opus 4.5 can now compress earlier parts of the conversation automatically, without notifying the user.

Read also: Google to retire Assistant in March 2026 as Gemini takes over

Anthropic says many of the upgrades are designed for complex “agentic” workflows in which Opus serves as a lead system orchestrating tasks handled by smaller, Haiku-powered agents. These scenarios require strong working-memory capabilities, an area the company claims has seen substantial gains in this release.

Opus 4.5 enters a competitive field. Other major AI developers have introduced flagship models in recent weeks, including OpenAI with GPT 5.1 on November 12 and Google with Gemini 3 on November 18.