Opus 4.5 in Claude Code
https://www.anthropic.com/news/claude-opus-4-5
Following Sonnet and Haiku, Anthropic's largest model has also been updated to ver 4.5. The problem with version 4 was its very high token price - now it's 3 times (!) cheaper. Therefore, the overall limits have been significantly increased, and Opus 4.5 has been added to Claude Code as an enhanced planning mode. It is capable of "creative problem-solving" — finding non-standard but legitimate solutions.

The main point is that the model uses significantly fewer tokens to achieve results, making it faster and cheaper to use.

Of course, in their programming and tool use tests, Opus 4.5 surpasses both Gemini 3 Pro and GPT-5.1. In one of the tests, the model handled a complex engineering task better than any human candidate. The model manages context and memory more effectively and can coordinate the work of several "sub-agents" to perform complex tasks.

https://www.youtube.com/watch?v=5O39UDNQ8DY

Claude Code is now available in a desktop application (research preview). You can run several local and remote Claude Code sessions in parallel: one agent fixes bugs, another explores GitHub, and a third updates documentation. Git worktrees are used for parallel work with repositories.

We are waiting for the model in most modern AI applications for code generation.

https://news.ycombinator.com/item?id=46037637
Several users emphasized that it's not the price per token that matters as much as the "cost per successful task." A smarter model, like Opus 4.5, makes fewer mistakes and requires fewer tokens to solve a problem, which can ultimately make it cheaper than "cheaper" but less intelligent models.

#newllmmodel #claudecode