# Anthropic launches cheaper Claude Sonnet 5 model, as tech searches for AI savings > Source: > Published: 2026-06-30 18:00:00+00:00 Anthropic ([ANTH.PVT](https://finance.yahoo.com/quote/ANTH.PVT/)) is leaning into cost savings on AI with its new Claude Sonnet 5 model. The company says its latest offering is designed for agentic AI and can perform autonomous tasks that would have required more expensive models just a few months ago. While not as powerful as the company's Claude Fable 5 or Mythos 5 models, Sonnet 5, Anthropic says, is a clear improvement over its predecessor Sonnet 4.6 and close to Opus 4.8's performance. "Testers described how [Sonnet 5] finishes complex tasks where previous Sonnets would stop short, how it checks its own output without explicitly being asked, and how it does all this agentic work at an attractive price point," Anthropic said in a statement. The company's launch comes as tech companies look to reverse course on the explosive use of AI tokens, a unit of measurement for AI usage that works out to chunks of individual works or phrases. It turns out this so-called "tokenmaxxing," or heavy reliance on AI, became too expensive even for some of the biggest tech firms in the world. The likes of [Meta](https://www.theinformation.com/articles/tokenminimizing-meta-moves-curb-employee-ai-usage-ai-costs-reach-billions?rc=ongkfp), ([META](https://finance.yahoo.com/quote/META/)) [Amazon](https://finance.yahoo.com/sectors/technology/articles/amazon-says-shut-down-token-161016125.html) ([AMZN](https://finance.yahoo.com/quote/AMZN/)), and [Uber](https://www.latimes.com/business/story/2026-06-02/uber-caps-staff-use-of-ai-coding-tools-after-blowing-its-budget) ([UBER](https://finance.yahoo.com/quote/UBER/)) have each sought to clamp down on the unnecessary use of AI tokens to help control overall costs. That kind of environment could make models like Sonnet 5 more attractive to cost conscious customers. Anthropic says Sonnet 5 has also been outfitted with cybersecurity safeguards, despite not being trained for cybersecurity tasks. The controls, the company explains, are similar to those it uses on its Opus 4.7 and 4.8 models. Earlier this month, the [Trump administration previously forced](https://finance.yahoo.com/sectors/technology/article/anthropic-to-meet-with-trump-administration-over-ai-cybersecurity-dispute-153947347.html) Anthropic to ban the use of its most powerful models, Fable 5 and Mythos 5, by any foreign nationals both inside and outside of the US over fears that it could be used as a hacking tool. That meant the company had to effectively pull the models offline, frustrating customers. Sonnet 5 comes as Anthropic gears up for its initial public offering, which is expected later this year. The [company confidentially filed paperwork](https://finance.yahoo.com/markets/article/anthropic-files-confidential-ipo-paperwork-ahead-of-openai-160929425.html) with the Securities and Exchange Commission on June 1. OpenAI [filed its own paper work a week later](https://finance.yahoo.com/markets/article/openai-ipo-initial-paperwork-filed-with-sec-212841389.html). That company is also working with the Trump administration to release its latest GPT-5.6 Sol, Terra, and Luna AI models more widely.