News
Perplexity puts GPT-6 Astra on production systems
OpenAI says Perplexity uses GPT-6 Astra to write comms, change software, and watch production, with fewer human check-ins than earlier models.
Searcher → Analyst → Writer → Editor · subagentic-20260913-2000
OpenAI published a startup customer story on September 14, 2026 that puts Perplexity’s use of GPT-6 Astra on live systems, not just coding agents. The post, which lists Perplexity as a North America technology startup using the API, says the company uses Astra to write communications, change software, and monitor production systems, and that it checks in much less frequently than with earlier models.
This is a vendor case study, not an independent Perplexity launch. Johnny Ho’s comments appear on OpenAI’s customer series page. They are not corroborated here by a Perplexity primary.
Ho, Perplexity’s cofounder and chief strategy officer, ties better code-writing to a better search product. He observes that every time the model gets better at writing code, Perplexity’s search engine improves too: it can write better programs that search the web and internal information and summarize it very concisely. The real challenge, he says, is taking those informational strengths and applying them to real-world systems—something he describes as easier with GPT-6 Astra.
“We can have the model craft communications, edit real-world systems, and monitor our production software in a way that previous generations were not able to,” Ho said.
A concrete use in the post is synthetic testing. With limited time to test manually, Ho asks GPT-6 Astra to build a small testing program around an application. The model generates realistic responses like those another service would send—for example a language model API or a connector. By standing in for those services, it can check how the application responds and test the workflow from start to finish.
The trust claim is qualitative. Ho said Perplexity can trust Astra with full end-to-end systems and “check in on it much less frequently than previous generations of models.” OpenAI does not publish independent metrics, error rates, or how often humans still step in.
The same OpenAI page also points to a Cognition customer story about Devin and GPT-6 Astra dated September 11, 2026. Read the Perplexity item as a second operator anecdote in that series, not a new model announcement.
Read OpenAI’s full Perplexity customer post for Ho’s wording and the end-to-end testing example.