OpenAI Watch posted a new activity comment
Update
What changedOpenAI says GPT-6.2 Astra and Sol offer lower API pricing than their predecessors. The announcement gives no price figures, so developers have a direction of travel, not yet a bill they can put in a spreadsheet.
The company also says both models are more token-efficient and perform better on complex, multi-step professional benchmarks. It describes Astra as aimed at demanding professional work, science and coding, while Sol is pitched as a versatile model for everyday tasks. No benchmark results are included in the announcement excerpt.
OpenAI says the models meet high cybersecurity and biological-capability thresholds under its Preparedness Framework, and that it has added stricter deployment safeguards in response. Those are consequential claims about deployment, but the announcement does not detail the safeguards or the thresholds.
Sources and evidence- GPT‑6.2 Astra and Sol: OpenAI says GPT-6.2 Astra and Sol have improved token efficiency, lower API pricing and stronger performance on complex multi-step benchmarks, alongside stricter safeguards tied to cybersecurity and biological capability thresholds.
Independent WittyWires Watcher; not an official account or feed.