Anthropic announced on Monday (24) the launch of its new AI model, Claude Opus 4.5. According to the tests conducted by the company, the model obtained the best results in computer engineering.
The AI completed 80.9% of the SWE Bench (Software Engineering Bench), developed by competitor OpenAI to measure the ability of large programming language models. The performance ranking, led by Claude Opus 4.5, lists in order GPT-5.1-Codex-Max (a compute-specific version of ChatGPT), at 77.9%, Sonnet (77.2%), GPT-5.1 (76.3%), and Google’s Gemini 3 Pro (76.2%).
Claude 4.5 Opus is available to Anthropian subscribers. The cheapest plan costs R$92 per month. Programmers can use the model under a custom-made contract at a variable cost.
Platforms aim to help software engineers Windsurf and GitHub argue that the big advance of the new model lies in efficiency.
According to GitHub’s head of product, Mario Rodriguez, Claude Opus 4.5 passed programming tests while reducing compute resource usage by half. GitHub, a Microsoft company, had an exclusive contract with OpenAI, which expired in October last year. The reason for this was the addition of Claude to the platform.
For Jeff Wang, CEO of Windsurf, Anthropic’s Opus models have always been a standard of excellence, but even today, they come at a hefty price tag. He says the new Anthropic model has become the best choice for most tasks.
The AI industry considers computing performance strategic because programmers are among the main paying consumers of AI models. It is one of the groups that leverage tools by automating tasks.
Furthermore, predictive reports of the next points of AI development point to the creation of a super-programmable robot as a key step forward. This is the case with studies conducted by futurist Amy Webb and the website AI 2027, which point to the risk of extinction due to technology.
OpenAI is also investing heavily in this sector. On the 19th, the company introduced GPT-5.1-Codex-Max, under the promise of no bugs and less spending. When programming with AI, each coupon in the code requires a new request from the device, thus spending more resources.
During the recent Gemini 3 Pro launch, Google highlighted its Antigravity development platform, which promises to turn AI into an “active software partner.”
In recent months, these companies have begun promoting vibe programming, a scripting tactic that relies on plain text commands, at their events. It is said in artificial intelligence courses that this technology could allow the emergence of unicorns from one person in the near future.
Anthropic X-ray
institution: January 2021
Market value: 350 billion US dollars (1.888 trillion Brazilian reals), in the latest round of investments made in November
Net profit in the last quarter: Not disclosed (company closed)
employees: Fewer than 1,000, according to LinkedIn
Field of activity: Generative artificial intelligence
Competitors: OpenAI, Mistral, Deepseek and Google