OpenAI announced the public release of its latest artificial intelligence model, GPT-5.6 Sol, on Thursday, marking its most advanced offering to date. Alongside the new model, the company introduced ChatGPT Work, an AI-powered assistant designed to streamline common office tasks by interacting with software applications such as spreadsheets, online calendars, and email platforms.
The unveiling intensifies competition in the AI sector, particularly with rival start-up Anthropic. Both companies’ latest models were closely scrutinized by the U.S. government due to national security and cybersecurity concerns. Last month, the administration issued an executive order seeking voluntary oversight over new AI models prior to public release, prompting extended discussions before ultimately allowing OpenAI and Anthropic to make their technologies broadly available.
Industry analysts note that OpenAI’s GPT-5.6 Sol delivers performance comparable to Anthropic’s current leading model, Fable 5. Vals A.I., which monitors AI capabilities, reported that in practical scenarios involving complex financial and legal tasks, OpenAI’s technology is state of the art. While OpenAI opted for a relatively open approach without extensive operational restrictions on Sol, Anthropic implemented tighter guardrails on Fable 5, aiming to prevent misuse related to cybersecurity vulnerabilities, biological information, and other sensitive areas.
Anthropic’s history with its AI systems highlights this cautious approach. In April, the company introduced Claude Mythos but did not release it publicly over security worries. Fable 5, a scaled-down version of Mythos, was launched last month but soon faced restrictions from the U.S. government, which barred foreign nationals from accessing it amid national security claims. In response, Anthropic temporarily took Fable offline and incorporated safeguards that reroute sensitive queries to a less powerful model, Claude Opus 4.8.
Both GPT-5.6 Sol and Fable specialize in identifying software vulnerabilities—a capability that poses a dual-use dilemma. While such technology can aid cybersecurity defenses for businesses and governments, it also risks being exploited by malicious actors. Anthropic’s protective measures limit offensive uses but may inadvertently hinder defensive applications. OpenAI, by contrast, maintains fewer limitations, intending to empower users in cybersecurity efforts.
Cost remains a consideration, as these advanced AI models come with premium pricing. Fable is approximately twice as expensive as Anthropic’s previous most powerful system, while OpenAI’s GPT-5.6 Sol can exceed Fable’s cost under certain conditions. Meta also introduced a new AI technology on Thursday, but it is less powerful than either OpenAI’s or Anthropic’s offerings and is priced notably lower.
The rollout of these advanced AI models occurs amid broader scrutiny of content use by AI developers. OpenAI and its partner Microsoft currently face litigation from a major news organization alleging copyright infringement related to AI training data, which both companies deny.