Anthropic announced the release of its latest artificial intelligence model, Claude Sonnet 4.5, on Monday, September 29th.

Anthropic claims the model offers enhanced capabilities in programming, computer usage, and meeting practical business needs, with particular strengths in specialized fields such as cybersecurity, finance, and scientific research. The Amazon-backed startup, valued at $183 billion, has now made Claude Sonnet 4.5 available to all users.

Anthropic claims that, based on industry benchmarks such as SWE-bench Verified (a test suite used to evaluate the software programming capabilities of AI systems), Claude Sonnet 4.5 is considered "the world's best programming model."

"As people use this model, they'll find it's smarter and more like a colleague," said Jared Kaplan, co-founder and chief scientific officer of Anthropic. "It's fun to collaborate with it, as we encounter problems and solve them together."

The company noted that Claude Sonnet 4.5 generates higher-quality code, is better at identifying areas for optimization, and is more reliable in following instructions.

Anthropic previously released Claude Opus 4.1 in August and Claude Sonnet 4 in May. The release of Claude Sonnet 4.5 further demonstrates the rapid pace of innovation in the AI ​​industry.

Anthropic was founded in 2021 by a group of former OpenAI researchers, and the two companies have been fierce competitors since their inception.

OpenAI launched the ChatGPT chatbot in 2022, sparking a wave of interest in generative AI. The startup's valuation has now soared to $500 billion, and it released its latest model, GPT-5, in August. However, the model's rollout has been bumpy, with some users complaining they can no longer use the company's older models.

Anthropic Chief Product Officer Mike Krieger stated that Claude Sonnet 4.5 will be the default model for users, and Anthropic recommends it as "suitable for nearly all use cases."

He also noted that other options remain: paid subscribers can still use the Opus model, and users with specific workflows who aren't ready to switch immediately can use the older Sonnet model.

Krieger added that while Claude Sonnet 4.5 is smaller than Claude Opus 4.1, it is smarter "in almost every way."

"We've found, and our customers have reported, that this model is incredibly useful in real-world situations," Krieger said.

Claude Sonnet 4.5 can operate autonomously for 30 hours, which Anthropic says allows it to remain focused on complex, multi-step tasks. By comparison, the company's Claude Opus 4, launched in May, has an autonomous runtime of just seven hours. Anthropic also claims that extensive security training has improved the model's behavior. The model's tendency to exhibit problematic behaviors, such as deception, control seeking, and flattery—saying only what the user wants to hear—has been significantly reduced.

Furthermore, Claude Sonnet 4.5 is more resilient to "cue injection attacks," which can trick a model into performing malicious actions, such as leaking sensitive data.

"I think this is probably the biggest breakthrough in model security we've seen in the last year to year and a half," Kaplan said.

Anthropic is already planning to release more new models. Kaplan revealed that even better-performing models are coming soon, "most likely including new versions of the Opus series."

"I can't promise anything right now," he said, "but I think we'll probably release one or two more models before the end of the year."