OpenAI announced a limited preview of its GPT-5.6 model family, including GPT-5.6 Sol, with access initially restricted during a cybersecurity review. Anthropic separately introduced Claude Sonnet 5, positioning it as a lower-priced model for agentic work in Claude products and its API.
OpenAI announced a limited preview of the GPT-5.6 family, including GPT-5.6 Sol, while Anthropic separately introduced Claude Sonnet 5 for broader workplace and developer use.
OpenAI said in its announcement, “Previewing GPT-5.6 Sol: a next-generation model,” that the GPT-5.6 family includes Sol, Terra, and Luna. According to OpenAI, the models are being previewed with improvements across coding, biology, cybersecurity, and reasoning modes.
OpenAI said GPT-5.6 availability will begin through the API and Codex for select partners. The company listed pricing from $1 to $5 per 1 million input tokens and $6 to $30 per 1 million output tokens, depending on the model or configuration.
The Associated Press reported that OpenAI’s GPT-5.6 Sol rollout is initially limited to government-approved customers during a cybersecurity review. AP described the restriction as applying to new AI models from OpenAI and Anthropic, with access limited during review for cybersecurity concerns.
That makes GPT-5.6 Sol a restricted preview rather than a general consumer launch. Based on OpenAI’s announcement and AP’s reporting, the main public details are the model family names, the focus areas for improvements, availability to selected partners, and the stated pricing bands.
Anthropic announced “Claude Sonnet 5,” describing it as an agentic Sonnet model with stronger reasoning, tool use, coding, and knowledge-work performance. The company said the model is available in Claude products and through the Anthropic API.
Anthropic listed introductory API pricing of $2 per 1 million input tokens and $10 per 1 million output tokens through August 31, 2026. The company positioned Claude Sonnet 5 as a model for everyday productivity and developer tasks rather than only high-risk or specialized deployments.
Axios also reported that Anthropic is presenting Claude Sonnet 5 as a lower-priced model intended to bring more agentic AI capabilities to everyday users. Axios said the model is not being treated as having the same cybersecurity risk profile as Anthropic’s strongest systems.
The announcements show different approaches to releasing advanced AI systems. OpenAI’s GPT-5.6 Sol preview is described as restricted, with AP reporting that access is limited to government-approved customers during cybersecurity review. Anthropic’s Claude Sonnet 5 is being offered in Claude products and the API, with public introductory pricing and an emphasis on everyday work.
Both companies are emphasizing reasoning, coding, and tool use, but the source material supports only limited comparisons. OpenAI says GPT-5.6 includes improvements in areas such as coding, biology, cybersecurity, and reasoning modes. Anthropic says Claude Sonnet 5 improves reasoning, tool use, coding, and knowledge-work performance.
For developers and organizations, the practical takeaway is that OpenAI’s GPT-5.6 Sol is not yet broadly available, while Anthropic’s Claude Sonnet 5 is being made available through product and API channels with published introductory pricing. Further performance claims will require independent testing beyond the companies’ launch statements and the available press reporting.
OpenAI announced a limited preview of the GPT 5.6 family, including GPT 5.6 Sol, while Anthropic separately introduced Claude Sonnet 5 for broader workplace and developer use.
OpenAI previews GPT 5.6 Sol with limited access OpenAI said in its announcement, “Previewing GPT 5.6 Sol: a next generation model,” that the GPT 5.6 family includes Sol, Terra, and Luna.
According to OpenAI, the models are being previewed with improvements across coding, biology, cybersecurity, and reasoning modes.
Continue reading
