OpenAI is retiring o3 from ChatGPT on 26 August 2026
OpenAI has set 26 August 2026 as the hard retirement date for o3 in ChatGPT, following a 90-day sunset period announced on 28 May 2026.
OpenAI announced on 28 May 2026 that o3 will be removed from ChatGPT on 26 August 2026, following a 90-day sunset period. If you use o3 as your preferred model in ChatGPT, you have a fixed deadline to plan around.
This change applies only to ChatGPT. The API is not affected.
What is actually happening
O3 is being phased out of ChatGPT as part of a broader programme of retiring older models that see limited ongoing usage. OpenAI’s stated rationale is straightforward: consolidating around fewer, more capable models lets the company direct compute and engineering resources towards its current flagship series.
O3 is not alone in this wave. GPT-4.5 is also being retired from ChatGPT, with a hard date of 27 June 2026 following a shorter 30-day sunset. Both models are currently available only to paid ChatGPT users via model settings.
This follows an earlier round of removals. On 13 February 2026, OpenAI retired GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT. The pattern is consistent: once a model family’s successor is established and performing well, the older versions are wound down in the consumer product.
What replaces o3
OpenAI is pointing users towards GPT-5.5, which is now the current flagship reasoning series in ChatGPT. The GPT-5.5 family includes GPT-5.5 Instant, GPT-5.5 Thinking, and GPT-5.5 Pro, and OpenAI describes the generation as delivering more accurate responses with improved writing style. Practically speaking, the updates focus on shorter, less bullet-heavy answers and more natural phrasing, which some users will find a welcome change from the more verbose output that characterised earlier reasoning models.
There is one thing worth clarifying: some reporting has suggested that existing o3 conversations will be automatically migrated to a GPT-5.5 variant. OpenAI’s release notes do not confirm this. The automatic continuation behaviour that has been documented applies to the GPT-5.1 line, where existing conversations moved to the corresponding GPT-5.3 or GPT-5.4 model. No equivalent automatic o3-to-GPT-5.5 migration is currently confirmed in official sources. If you have ongoing o3 conversations you want to continue, the safe assumption is that you will need to switch models manually before 26 August 2026.
What does this mean for you
For most paid ChatGPT users, the practical impact depends on how much you have been relying on o3 specifically.
O3 was OpenAI’s main reasoning model before the GPT-5.x series arrived. It uses reinforcement learning to work through a private chain of thought before generating a response, which made it well suited to complex maths, science, and coding tasks. If those are your use cases, GPT-5.5 Thinking or GPT-5.5 Pro are the natural successors. GPT-5.5 Pro, like o3-pro before it, is designed to spend more time reasoning through problems before responding, and it has access to the same tools: web search, file analysis, Python, visual reasoning, and memory.
If you are a lighter user who picked o3 because it happened to be in the model list, switching to GPT-5.5 Instant or letting ChatGPT’s Auto routing handle model selection will serve you well without any noticeable disruption.
For paid users who hit rate limits on GPT-5.4 Thinking, GPT-5.4 mini is the designated fallback. Enterprise customers retain the option to default Auto routing to GPT-5.4 mini if they prefer consistency over capability.
A note for developers
While the ChatGPT retirement is the news here, there is a related deadline worth noting separately. On 11 June 2026, OpenAI notified developers that six of the earliest GPT-5 and o3 model snapshots will stop responding to API requests on 11 December 2026. If any routing policy or production integration references those date-stamped model IDs, you have until December to update them. The OpenAI deprecations page has the full list of affected snapshot identifiers.
Key dates at a glance
- 27 June 2026: GPT-4.5 retired from ChatGPT
- 26 August 2026: O3 retired from ChatGPT
- 11 December 2026: Early GPT-5 and o3 API snapshots stop responding
If you are a paid ChatGPT user who actively selects o3, now is a good time to spend a few sessions with GPT-5.5 Thinking and see how it handles your typical workload. The transition is unlikely to be jarring, and in most cases the newer model should simply be better at the tasks that made o3 worth using in the first place.