OpenAI realized from the experiment, Altman stated, with early testers demanding: “make it cheaper and make it sooner.” The reply to that’s GPT 6.1 Sol, which offers close to Astra stage intelligence at one-fifth of the value. Cached inputs value $0.10 (US) per million tokens, 95% lower than commonplace enter pricing. This makes it cheaper for brokers to work by way of longer duties, reuse context, and ruminate on options. Instance makes use of embody multi-step enterprise workflows, code writing and debugging, and doc evaluation.
GPT 6.1 Sol is “smarter than Astra in some methods, and it’ll turn into your day by day workhorse,” Altman stated. It’s now accessible to all Plus, Professional, Enterprise, Enterprise, and Edu customers of ChatGPT Work and Codex, and can quickly be supplied in an ‘Ultrafast’ plan, with as much as 8X sooner token technology.
Levy famous that GPT 6.1 Sol reduces the associated fee curve for enterprises seeking to deploy AI brokers at scale, all whereas “avoiding among the privateness/security/integrity baggage of the now-stillborn GPT 6.1 Astra.” It successfully shifts the economics of AI brokers by dropping the value of resource-intensive growth actions like querying repositories, working checks, and iterating code.
