OpenAI Launches GPT-6.2 Astra and Sol Amid Tough Early Reviews
OpenAI released its GPT-6.2 Astra and Sol model upgrades to mixed reactions, landing just ahead of rival Anthropic's public stock offering.

OpenAI has officially launched its latest artificial intelligence models, GPT-6.2 Astra and the more budget-friendly GPT-6.2 Sol, designed specifically to handle professional workflows. However, the rollout has been met with harsh feedback from early testers who feel the software struggles to keep pace with competing products in the market.
According to independent benchmarks from Artificial Analysis, Astra has climbed five spots since September, placing it directly behind Fable 5.2 and ahead of Google's Gemini 4 Carbon. Meanwhile, the economical Sol model matches the performance scores of the older GPT-6 Astra while reducing task execution costs by roughly 15%. Although Astra performs strongly in scientific reasoning, legal tasks, robotics, and 3D modeling, it shows no meaningful improvements in coding capabilities. Reviewers were notably critical, with some labeling the release poorly and pointing out readability issues in the generated answers.
What it means for developers
Developers evaluating the new OpenAI lineup will notice a mixed set of capabilities, particularly with coding performance remaining stagnant or even regressing in the Sol variant. Choosing the right infrastructure matters heavily for cost and output quality. For those looking to experiment, developers can try top AI models cheaply through one API at https://apixoai.online without needing multiple vendor subscriptions.
Cybersecurity partnerships and wider industry changes
In related security news, OpenAI disclosed that NATO's technology and cyber agency has deployed its Daybreak models to help defend allied networks, processing up to 300 billion tokens per month. Additionally, European cybersecurity agency ENISA utilized the models to uncover four distinct software vulnerabilities in European Union institutions, all of which have since received security patches. Meanwhile, independent security researchers using AI tooling successfully uncovered a critical Java vulnerability affecting widespread enterprise systems, prompting emergency fixes from Oracle.
Source: OpenAI’s newest release lands to rough reviews — The Rundown AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key

