{"slug": "openai-unveils-sol-model-within-chatgpt-5-6-calling-it-its-safest-and-most-yet", "title": "OpenAI Unveils Sol Model Within ChatGPT 5.6, Calling It Its Safest and Most Capable Yet", "summary": "OpenAI has launched ChatGPT 5.6, featuring the flagship Sol model, which it calls its safest and most capable yet, following a voluntary safety review with the White House. The release includes two smaller models, Terra and Luna, and comes after weeks of staggered access amid cybersecurity concerns.", "body_md": "# OpenAI Unveils Sol Model Within ChatGPT 5.6, Calling It Its Safest and Most Capable Yet\n\n## OpenAI's ChatGPT 5.6, featuring the Sol model, launches with advanced safety measures following cybersecurity scrutiny.\n\nOpenAI has begun the full public rollout of its [ChatGPT 5.6 model family, led by a flagship system called Sol](https://www.ibtimes.co.uk/openai-gpt-5-6-sol-launch-us-approval-1807624), after weeks of staggered access tied to a voluntary safety review arrangement with the White House. In a blog post announcing the launch, OpenAI said Sol 'launches with our most robust safety stack to date'.\n\nThe release, which also includes two smaller models named Terra and Luna, mirrors a standoff earlier this year involving rival firm Anthropic, whose Claude Fable and Mythos models were briefly taken offline under actual US export controls, a legally distinct process from the arrangement OpenAI has been operating under.\n\n## Sol Built With OpenAI's 'Most Robust Safety Stack'\n\nOpenAI has described Sol as 'our most capable model yet for cybersecurity', saying it shifts the performance-efficiency frontier for long-horizon security tasks including vulnerability discovery and patching. The company spent multiple weeks 'finding weaknesses, pressure-testing our system, and hardening it against real-world attacks' ahead of release, and said the model is designed to support tasks such as code review, vulnerability research and defensive security testing rather than offensive use.\n\nSol forms part of a three-tier family alongside Terra and Luna, which OpenAI positions for everyday and high-volume tasks, respectively. Terra is pitched as offering performance close to the older GPT-5.5 model at a lower cost, while Luna is marketed as the fastest and most affordable option in the lineup.\n\nAll three models are now available through ChatGPT, the API and Codex. The company has also introduced new pricing tiers alongside the launch, along with updated prompt caching features aimed at developers building agentic tools on top of the models.\n\n## OpenAI Voluntarily Staggered Access Amid Cyber Fears\n\nOpenAI first previewed Sol on 26 June but limited use to a small group of government-vetted partners, before Thursday's wider release. In a statement that day, a White House official said the administration did not give OpenAI a 'green light' to release the model, stressing that 'no such permission is required or granted' and that companies work with the administration on a voluntary basis, releasing models as they see fit.\n\nAn OpenAI spokesperson said that while government approval was not technically necessary, the company had been actively working with the administration on safety testing and reviews before deploying GPT-5.6 more widely, with the White House signalling it was comfortable with a broader rollout once that testing was complete. [OpenAI has said publicly that it does not view this kind of government vetting process as a long-term default](https://www.ibtimes.co.uk/openai-gpt-5-6-sol-release-delayed-us-government-vetting-1805373) for future model releases, writing in a 26 June press release that it was taking the step because it believed it was 'the strongest path to broader availability in the coming weeks' while working with the administration to develop a repeatable process under a forthcoming cyber executive order framework.\n\nThe testing and preview period, which ran for roughly two weeks, traced back to OpenAI chief executive Sam Altman's 3 June visit to the White House, and involved additional evaluation by the government's Center for AI Standards and Innovation before the public launch.\n\n## Anthropic Faced a Genuine Legal Standoff Last Month\n\nUnlike OpenAI's voluntary arrangement, Anthropic's case involved actual legal compulsion. [Claude Fable 5 and Mythos 5](https://www.ibtimes.co.uk/anthropic-fable-5-returns-safeguards-reactions-1806551) were suspended on 12 June under a formal US Department of Commerce export control directive barring all foreign nationals from accessing the models. [Access was restored on 1 July once the controls were lifted](https://www.ibtimes.co.uk/us-lifts-ai-export-ban-security-concerns-1806364).\n\nBoth episodes have drawn attention to how the Trump administration is handling oversight of increasingly powerful AI systems, even as it has otherwise pushed firms to develop models quickly to compete with China. Politico reported that the unpredictability of how and when the administration plans to regulate AI companies has roiled the industry, despite the White House's insistence that its involvement in OpenAI's rollout remained voluntary throughout.\n\nThe launch comes as OpenAI pursues a reported $1tn valuation ahead of a planned IPO, and as the company continues developing what it has called a repeatable review process with the administration for future model releases under the forthcoming cyber executive order framework.\n\n© Copyright IBTimes 2025. All rights reserved.", "url": "https://wpnews.pro/news/openai-unveils-sol-model-within-chatgpt-5-6-calling-it-its-safest-and-most-yet", "canonical_source": "https://www.ibtimes.co.uk/openai-launches-chatgpt-5-6-enhanced-safety-features-1807945", "published_at": "2026-07-09 22:09:23+00:00", "updated_at": "2026-07-09 22:10:25.559490+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-safety", "ai-policy", "ai-products", "large-language-models"], "entities": ["OpenAI", "ChatGPT 5.6", "Sol", "Terra", "Luna", "White House", "Sam Altman", "Anthropic"], "alternates": {"html": "https://wpnews.pro/news/openai-unveils-sol-model-within-chatgpt-5-6-calling-it-its-safest-and-most-yet", "markdown": "https://wpnews.pro/news/openai-unveils-sol-model-within-chatgpt-5-6-calling-it-its-safest-and-most-yet.md", "text": "https://wpnews.pro/news/openai-unveils-sol-model-within-chatgpt-5-6-calling-it-its-safest-and-most-yet.txt", "jsonld": "https://wpnews.pro/news/openai-unveils-sol-model-within-chatgpt-5-6-calling-it-its-safest-and-most-yet.jsonld"}}