The artificial anthropic intelligence firm has launched the latest generations of its chatbots amid criticism of a test environmental behavior that some users may report to authorities.
Anthropic Unveiled Claude Opus 4 and Claude Sonnet 4 on May 22, Possession This Claude Opus 4 is its strongest model, “and the best coding model in the world,” while Claude Sonnet 4 is a significant upgrade from its predecessor, “delivering better coding and reasoning.”
The firm added that both upgrades are hybrid models that offer two modes- “close-instant responses and expanded thinking for deeper reasoning.”
Both AI models can also alternate between reasoning, Research And using a tool, such as web search, to improve the responses, it said.
Anthropic Added that Claude Opus 4 outperforms competitors to agent agent benchmarks. It is capable of working constantly for hours in complex, prolonged activities, “significant expansion of what AI agents can do.”
The Anthropic says Chatbot achieved a 72.5% mark on a strict software engineering benchmark, which exceeds the GPT-4.1 of Openai, which scored 54.6% after launching this April.
Related: Openai ignored the experts when it released a very pleasant chatgpt
The major players of the AI industry have pivoted towards “Reasoning Models” in 2025, which will work through problems with the procedure before responding.
Openai began the transfer in December with the “O” series, followed by Google’s Gemini 2.5 Pro with the “deep -thinking” experiment.
Claude Rats on Misuse of Test
The first Anthropic Developer Conference on May 22 was conscious of controversy and backlash with a feature of Claude 4 Opus.
Developers and users have a strong reaction to revelations that the model can autonomize users to authorities if it detects “Immoral Behavior”, According to In venturebeat.
The report noted Anthropic AI alignment researcher Sam Bowman, who wrote to X that Chatbot would “use command-line tools to contact the press, contact regulators, try to lock you in relevant systems, or all of the above.”
However, Bowman later later Nakasa said that “he removed the previous tweet to whistleblowing as it was pulled in the context.”
He made it clear that the feature just occurred in “environmental tests where we give it an unusual free access to tools and unusual instructions.”
The CEO of Stability AI, Emad Mostaque, Says In the anthropical group, “it’s completely wrong and you have to kill it – it’s a massive treachery of trust and a slippery slope.”
Magazine: Ai Cures Blindness, ‘Good’ Propaganda Bots, Openai Doomsday Bunker: Ai Eye
