Parallel says GPT-6 Astra halved both time and cost on a multi-state labor-market research agent task
OpenAI Blog·low signal
In an OpenAI customer post, Parallel Web Systems says Astra finished a task covering six labor-market statistics across four states and six months in half the time and at about half the cost of prior models. The saving came from fewer research calls, fewer tokens, more targeted queries and delegation to parallel sub-agents. It is a vendor case study that does not name the baseline model, so treat the 2x as a claim, not a benchmark.