Perplexity Trusts GPT-6 Astra with Full End-to-End Systems

OpenAI published a customer story about how Perplexity uses the GPT-6 Astra model to write communications, modify real software systems, monitor production, and perform end-to-end testing with simulated external services.

Compiled by Sergey KostenchukPublished 2026-09-14Updated 2026-09-18

2026-09-14 Business OpenAI

💼 Perplexity Trusts GPT-6 Astra with Full End-to-End Systems

OpenAI published a customer story: Perplexity uses GPT-6 Astra (released on September 3, 2026) to write communications, modify real software systems, and monitor production. Perplexity co-founder Johnny Ho stated that the team can already trust the model with full systems and checks it less frequently than previous generations.

🌍 The second published GPT-6 Astra case study: agentic models are transitioning from review hints to multi-step production work with less human oversight. This changes the requirements for permissions, verification, and rollback when modifying live systems.

👤 Astra builds its own tests and simulates external services (LLM APIs, connectors) for end-to-end runs without live dependencies. There are no independent benchmarks in the case study — reliability claims are currently self-assessments by Perplexity and OpenAI.

Source 1: https://openai.com/index/perplexity-improving-accuracy-with-astra

Sources