Please, OpenAI and Anthropic, stop releasing new models. At this rate, I’m about to become a fluff piece myself.
OpenAI claimed yesterday that for safety reasons—out of a profound sense of responsibility for humanity’s future and to prevent an out-of-control AI from destroying humanity (I added these two parts)—it is pausing the release of the latest Astra. These days, if a model doesn’t have safety risks, it’s embarrassed to call itself a cutting-edge frontier model.
Right as Anthropic just released Claude Sonnet 5.5 and Opus 5.5 to seize the productivity market, OpenAI turns around and fights in-your-face with 6.1 Sol.
In the DeepSWE software engineering benchmark, 6.1 Sol matched Astra at one-fifth the cost, up 6.4 percentage points over the previous-generation Sol. On the complex document benchmark GDP.pdf and the multi-step business workflow AutomationBench, it beats Opus 5.5 across the board.
On pricing, standard input is $2 per million tokens, output $10. API prices are this cheap now, so deducting 50% of the quota from subscriptions is reasonable, right?
At the start of the month, they were still signing joint letters calling for legislation to set limits and slow the pace of R&D; by the end of the month, new models iterate weekly, and the top flagship Astra continues to live in the safety narrative, serving as a moral totem.
Those who got ashore first are eager to pull up the ladder; those who got ashore later are eager to flip the table. Who will flip the table next? Gemini?