Key points
- OpenAI released GPT-6.2 Astra (frontier model) and GPT-6.2 Sol (everyday model) to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, plus the API.
- Astra API rates fell 20% to $8 input, $0.40 cached input and $40 output per million tokens; Sol keeps $2, $0.10 and $10.
- Artificial Analysis ranks Astra second on its Intelligence Index behind Claude Fable 5.2, and reports flat coding scores and high hallucination rates on hard factual questions.
- Astra is trained to follow slide templates, down to font size, and to match professional conventions in law, finance and sales.
Why it matters for marketers: Marketers get cheaper frontier-model access and better template-following for decks and documents, but independent tests show uneven results in coding, legal work and factual accuracy, so pilot before switching.
OpenAI has released GPT-6.2 Astra, aimed at its hardest professional work, and GPT-6.2 Sol, an efficient model for everyday tasks. Both are available now in ChatGPT Work and Codex for paid individual, Enterprise and Edu plans, and through the API. Astra's API prices dropped 20% from GPT-6 Astra, while Sol's prices are unchanged from GPT-6.1 Sol.
For content teams, the most relevant claim is output quality. OpenAI says Astra produces slides, spreadsheets and documents that read less like AI, follow supplied templates more closely, and use titles that state the main takeaway. It says the model can work from an original file or a screenshot and match layout, colors and typography. OpenAI also says Sol comes within 0.5 points of Astra on one coding benchmark at about one-fifth the cost.
Independent results are more mixed. According to Artificial Analysis, Astra ranks second on its Intelligence Index behind Claude Fable 5.2, but leads on cost-performance. The firm also reports no coding gains and hallucination rates of 59% for Astra and 55% for Sol on very hard factual questions. Vals AI found the models rank lower on legal benchmarks than its overall index. Latent Space noted sharply negative reactions from some influencers.
OpenAI also says both models meet its Critical threshold for cybersecurity, so it has tightened safeguards, which it says limit some legitimate security testing. Teams should run their own side-by-side tests on real templates and briefs, check cost per task rather than list price, and keep human review on factual claims.
Sources
This brief is our own summary. Read the full reporting at the source.
- OpenAI: Introducing GPT-6.2 Astra and Sol
- The Decoder: OpenAI launches GPT-6.2 Astra and Sol with lower API prices and some benchmark gains
- Latent Space: [AINews] GPT-6.2 Astra, the #2 Frontier Model is better and 20%-600% cheaper, but has mixed reviews; 6.2 Sol likely main workhorse
- TLDR AI: GPT-6.2 Astra 🧠, Microsoft’s seven models 🚀, Replit Missions 🤖
- MarkTechPost: OpenAI Releases GPT-6.2 Astra and Sol With a 20% Cheaper Astra API