OpenAI launches GPT-5 as its new default model
OpenAI released GPT-5, unifying its reasoning and fast models behind a single system that decides how hard to "think" per request. It's available in ChatGPT for all users (including free) and via the API, and OpenAI positions it as their most capable model for coding, writing, and agentic tasks — with significantly lower hallucination rates than GPT-4o and o3.
The gap between "the model your ops team uses in ChatGPT" and "the model you build automations on" is basically closed — same brain, same reliability.
If you were waiting for AI to be "reliable enough" for customer-facing logistics use cases (draft replies, quote summaries, incident notes), the hallucination drop is the number that matters. Re-run the pilots you shelved 6 months ago on GPT-5 before assuming the answer is still "not yet".
https://openai.com/index/introducing-gpt-5/