Ahropic on Wednesday released a new version of its lightweight AI model Claude Haiku 4.5 unveiled According to the company, the new version is similar in performance to the Sonnet 4 model, but with One third of the cost And over Twice the speed It works.
Claude Haiku 4.5 AI performance in benchmarks
According to Eropic’s iernal tests, the Claude Haiku 4.5 model has succeeded in the SWE-Bench verified benchmark. 73 perce and earn pois in the Terminal-Bench (focused on the command line). 41 perce has taken However, although its performance is slightly weaker than the Sonnet 4.5, it is on par with the GPT-5 and Gemini 2.5 models.
Similar results were recorded in other benchmarks such as using tools, working with computers and visual reasoning.
New Haiku model from today for all users Free plans Ahropic is available. The company says Haiku 4.5 will be particularly attractive to free versions of its AI products, as it offers impressive capabilities while putting little strain on servers. The lightness of the model makes it possible to run multiple Haiku ages simultaneously or in combination with more advanced models.
Mike Krieger, Ahropic’s chief product officer, said in a stateme about the model:
“This model opens the gate of a new generation of practical applications of artificial ielligence; Where Sonnet is responsible for complex programming and Haiku-based ages run at high speed. We now have a complete toolbox of ages, each model offering a specific combination of ielligence, speed and cost for differe tasks.”
According to Ahropic, the immediate applications of Haiku 4.5 are likely to be in Software developme tools will be seen
