Anthropic has acknowledged that its Opus 4.7 model is less capable than the upcoming Claude Mythos Preview, particularly in cybersecurity applications. The admission comes as the White House prepares government agencies to adopt Mythos.
Anthropic disclosed that Opus 4.7 represents a step backward compared to Mythos Preview across general capabilities and cyber skills, according to reporting from CNBC's Ashley Capoot.
The statement underscores the significance of Mythos, which has drawn attention from early testers who describe it as a "potent" tool for cybersecurity work. Those with access to the model report sharp improvements over Anthropic's prior offerings.
Meanwhile, the White House is moving quickly to integrate Mythos into government operations. A memo from the Office of Management and Budget (OMB) has notified Cabinet departments that protections are being established to allow federal agencies to begin using Mythos, according to Bloomberg.
The setup suggests the administration views Mythos as sufficiently secure and valuable for sensitive government work. The OMB's proactive framework indicates coordination between Anthropic and federal leadership on deployment safeguards.
Anthropic's candor about Opus 4.7's limitations is notable. Rather than positioning it as an incremental improvement, the company has been direct about where the newer model falls short. This transparency may signal confidence in Mythos's capabilities or reflect internal assessments about which product deserves institutional backing.
The timing aligns with broader government interest in AI adoption. Federal agencies have been cautious about AI deployment but are under pressure to modernize. Mythos's cybersecurity strengths appear to have convinced the White House that structured access is warranted.
Anthropologic's Mythos remains in preview status, meaning wider adoption will likely follow initial government testing. The OMB protections framework will be critical in determining how quickly federal agencies can operationalize the model for real work.
Z.ai released GLM-5.3's weights on Hugging Face under a new license that requires large companies to undergo security review before hosting the model. The change marks a departure from the standard MIT license.
Anthropic has introduced the Model Hardware Standard (MHS), a unified interface enabling AI agents to operate robotic arms, lab instruments, and other physical devices. Early testing shows integration time has dropped from weeks to hours.
Open-weight AI companies—those releasing freely available models—are attracting major acquisition interest from tech giants. The trend reflects growing capital investment in the business model of distributing AI models at no cost.
Google Deepmind has upgraded its Co-Scientist AI system to autonomously plan experiments, operate lab equipment, and publish scientific papers. The Gemini-based multi-agent platform demonstrated experimentally validated results across materials science, chemistry, and medical AI development.