What Changed in the Latest OpenAI Model Release? (Oct 2026)
The short answer
OpenAI’s latest model release, on October 7, 2026, brought GPT-6 to everyone who uses ChatGPT. Paid plans (Plus, Pro, Business, Enterprise) got it that day, powered by GPT-6 Sol; Free and Go followed on October 8 with GPT-6 Luna. Two features came with it: Intelligent UI — answers with interactive charts, buttons, forms and small tools — and answers that start while the model is still thinking. ChatGPT Work and Codex models did not change, and API prices did not move.
What changed, point by point
| Change | Detail (October 7–8, 2026) |
|---|---|
| Default chat model | GPT-6 in the Chat tab; replaces GPT-5.6 for chat |
| Paid plans | GPT-6 Sol (Plus, Pro, Business, Enterprise; Enterprise subject to admin settings) |
| Free and Go | GPT-6 Luna, from October 8 |
| Pro reasoning | Still GPT-6 Astra; no Intelligent UI |
| Intelligent UI | Interactive components rendered progressively as the model generates |
| Thinking + answering | Answer streams in parts while reasoning continues |
| Web search | GPT-6 Instant starts answering 44% sooner than GPT-5.6 Instant on searched questions (OpenAI) |
| Work and Codex | Unchanged in this release |
| API prices | Unchanged |
The two new behaviours
Intelligent UI. OpenAI trained GPT-6 to decide how to present an answer: side-by-side for comparisons, an interactive diagram for an explanation, a map for a road trip, plain text when that is best. It can also build a working tool on request, such as a savings calculator or bill splitter. Under the hood, a library of native streamable components and a compiler render the interface while the model is still generating.
Answering while thinking. Reasoning models used to make you wait for the whole chain of thought. GPT-6 composes the answer across partial responses as it works. OpenAI says GPT-6 at Extra High effort begins answering in the same time as GPT-5.6 at Medium while scoring better than GPT-5.6 at Extra High on its internal agentic-task evaluation.
Safety notes
OpenAI says GPT-6 builds on GPT-6 Astra’s safety work, with stronger resistance to multi-turn jailbreaks and clearer statements of what it cannot do. Its system card for the October models also reports statistically significant regressions on some disallowed-content categories against the GPT-5.6 versions — self-harm for Sol, and self-harm, gore and sexual content for Luna. Teams deploying ChatGPT to minors or in sensitive settings should read it.
What this means for developers
Nothing changes in your API code. The ChatGPT models map to API models you can already call: gpt-6.1-sol ($2/$10 per million tokens, $0.10 cached input), gpt-6-luna ($0.10/$0.50) and gpt-6-astra ($10/$50) — see current API prices. One related change did land in Codex: GPT-5.5 was retired from Codex sessions that sign in with ChatGPT on October 14, 2026; API-key sessions are unaffected.
If you are choosing between the three GPT-6 tiers for a workload, see GPT-6 Sol vs Luna vs Astra. The September launch details are in every OpenAI DevDay 2026 announcement.
Last verified: October 15, 2026.