For users who still rely on these models for specific workflows, there is a brief window to transition. Here is what is happening, why, and what it means for both ChatGPT users and API developers.

The Final Sunset Dates: When Access Ends

OpenAI structured the removal with two different sunset periods, giving more time for the more distinct o3 reasoning model.

• GPT-4.5 — Enters a 30-day transition period and will be shut down from ChatGPT on June 27, 2026 .
• OpenAI o3 — Enters a 90-day sunset period and will be shut down from ChatGPT on August 26, 2026 .

These models were not available to free users. Access was limited to paid ChatGPT subscribers on Plus, Pro, Team, and Enterprise plans, who could enable them through a manual "additional models" toggle in the settings . The removal applies across both the web and mobile ChatGPT interfaces, though Enterprise workspaces may have a brief grace period beyond the public dates .

Crucially, this sunset is confined to the consumer chat product. "These changes apply to ChatGPT only; there are no changes to the API," the release notes explicitly state . Developers using these models via the API can continue to do so on a separate, slower deprecation timeline.

A Year of Aggressive Phaseouts: The Full GPT-4 Purge

This isn't a sudden decision but the final scene in a year-long strategy of rapid model consolidation. OpenAI initially planned to retire many of these models in August 2025 alongside the launch of GPT-5 but temporarily reversed course on GPT-4o after vocal backlash from Plus and Pro users who preferred its conversational warmth for creative tasks .

However, the reprieve was short-lived. By early 2026, citing usage rates as low as 0.1% for GPT-4o, the company resumed its aggressive pruning . The resulting timeline shows a steady march toward a single-model ecosystem:

• August 2025: Plans are first announced to retire GPT-4o, GPT-4.1, GPT-4.5, and others alongside the GPT-5 launch, later partially walked back .
• February 13, 2026: The major cut happens. GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini are retired alongside GPT-5 (Instant and Thinking) .
• March 11, 2026: The next layer of the stack is removed as ChatGPT retires the GPT-5.1 Instant, GPT-5.1 Thinking, and GPT-5.1 Pro models .
• June–August 2026: The final GPT-4 generation model (GPT-4.5) and the last legacy o-series reasoning model (o3) are removed, leaving no pre-GPT-5.5 models selectable in the interface .

With the removal of GPT-4.5, no GPT-4-era model remains available in ChatGPT. It is a symbolic and technical milestone for the platform .

Why OpenAI Is Consolidating Around GPT-5.5

OpenAI’s stated reasoning is straightforward: to "better serve our newer, most capable models" . The company has been unifying its product and API around the GPT-5.5 family, which includes GPT-5.5 Instant and GPT-5.5 Thinking, both released on April 24, 2026 .

This consolidation strategy serves three main goals:

1. Simplify the Model Picker: A sprawling list of models, many with similar names and subtle differences in behavior, created confusion for non-technical users. Slimming down the options is a direct attempt to reduce that friction and lower the support burden .
2. Improve Reliability and Focus Resources: Maintaining production infrastructure for a long tail of legacy models with minuscule usage is inefficient. OpenAI revealed that GPT-4o was used by less than 0.1% of users daily, a key justification for its removal . Shifting server capacity to the newer GPT-5.x line allows for faster iteration and better performance for the models that almost everyone is actually using .
3. Unify the User Experience: By forcing all conversations to a modern model family, OpenAI also retires associated legacy features. The release notes specifically mention retiring the "canvas" interface from GPT-5.5 models, with its functionality folded directly into new, standard chat-based writing and coding blocks .

A Different Clock for Developers: The API Deprecation Timeline

The API exists on a completely separate schedule. While the consumer product has been ruthlessly streamlined, developers have a much longer runway for migrating their applications. Key upcoming dates on the API calendar include:

• July 14, 2025: gpt-4.5-preview was shut down, with gpt-4.1 recommended as a replacement .
• March 26, 2026: Several legacy GPT-4 API snapshots, including gpt-4-0314 and gpt-4-1106-preview, are scheduled for shutdown .
• July 23, 2026: A large batch of preview models including computer-use-preview, gpt-4o-audio-preview, and gpt-4o-search-preview will be turned off .
• August 10, 2026: The gpt-5.2-chat-latest and gpt-5.3-chat-latest snapshots will be removed, with gpt-5.5 as the direct replacement .
• October 23, 2026: The o3-mini-2025-01-31 model and other legacy models like gpt-3.5-turbo-0125 will reach end of life, with gpt-5.5 as the recommended migration path .

The pattern is clear: the API is on a longer, more methodical, staggered schedule, while the consumer-facing ChatGPT has been used as the sharp point of the spear to push users into the latest generation of models as quickly as possible .

For users who still rely on these models for specific workflows, there is a brief window to transition. Here is what is happening, why, and what it means for both ChatGPT users and API developers.

The Final Sunset Dates: When Access Ends

OpenAI structured the removal with two different sunset periods, giving more time for the more distinct o3 reasoning model.

• GPT-4.5 — Enters a 30-day transition period and will be shut down from ChatGPT on June 27, 2026 .
• OpenAI o3 — Enters a 90-day sunset period and will be shut down from ChatGPT on August 26, 2026 .

These models were not available to free users. Access was limited to paid ChatGPT subscribers on Plus, Pro, Team, and Enterprise plans, who could enable them through a manual "additional models" toggle in the settings . The removal applies across both the web and mobile ChatGPT interfaces, though Enterprise workspaces may have a brief grace period beyond the public dates .

Crucially, this sunset is confined to the consumer chat product. "These changes apply to ChatGPT only; there are no changes to the API," the release notes explicitly state . Developers using these models via the API can continue to do so on a separate, slower deprecation timeline.

A Year of Aggressive Phaseouts: The Full GPT-4 Purge

This isn't a sudden decision but the final scene in a year-long strategy of rapid model consolidation. OpenAI initially planned to retire many of these models in August 2025 alongside the launch of GPT-5 but temporarily reversed course on GPT-4o after vocal backlash from Plus and Pro users who preferred its conversational warmth for creative tasks .

However, the reprieve was short-lived. By early 2026, citing usage rates as low as 0.1% for GPT-4o, the company resumed its aggressive pruning . The resulting timeline shows a steady march toward a single-model ecosystem:

• August 2025: Plans are first announced to retire GPT-4o, GPT-4.1, GPT-4.5, and others alongside the GPT-5 launch, later partially walked back .
• February 13, 2026: The major cut happens. GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini are retired alongside GPT-5 (Instant and Thinking) .
• March 11, 2026: The next layer of the stack is removed as ChatGPT retires the GPT-5.1 Instant, GPT-5.1 Thinking, and GPT-5.1 Pro models .
• June–August 2026: The final GPT-4 generation model (GPT-4.5) and the last legacy o-series reasoning model (o3) are removed, leaving no pre-GPT-5.5 models selectable in the interface .

With the removal of GPT-4.5, no GPT-4-era model remains available in ChatGPT. It is a symbolic and technical milestone for the platform .

Why OpenAI Is Consolidating Around GPT-5.5

OpenAI’s stated reasoning is straightforward: to "better serve our newer, most capable models" . The company has been unifying its product and API around the GPT-5.5 family, which includes GPT-5.5 Instant and GPT-5.5 Thinking, both released on April 24, 2026 .

This consolidation strategy serves three main goals:

1. Simplify the Model Picker: A sprawling list of models, many with similar names and subtle differences in behavior, created confusion for non-technical users. Slimming down the options is a direct attempt to reduce that friction and lower the support burden .
2. Improve Reliability and Focus Resources: Maintaining production infrastructure for a long tail of legacy models with minuscule usage is inefficient. OpenAI revealed that GPT-4o was used by less than 0.1% of users daily, a key justification for its removal . Shifting server capacity to the newer GPT-5.x line allows for faster iteration and better performance for the models that almost everyone is actually using .
3. Unify the User Experience: By forcing all conversations to a modern model family, OpenAI also retires associated legacy features. The release notes specifically mention retiring the "canvas" interface from GPT-5.5 models, with its functionality folded directly into new, standard chat-based writing and coding blocks .

A Different Clock for Developers: The API Deprecation Timeline

The API exists on a completely separate schedule. While the consumer product has been ruthlessly streamlined, developers have a much longer runway for migrating their applications. Key upcoming dates on the API calendar include:

• July 14, 2025: gpt-4.5-preview was shut down, with gpt-4.1 recommended as a replacement .
• March 26, 2026: Several legacy GPT-4 API snapshots, including gpt-4-0314 and gpt-4-1106-preview, are scheduled for shutdown .
• July 23, 2026: A large batch of preview models including computer-use-preview, gpt-4o-audio-preview, and gpt-4o-search-preview will be turned off .
• August 10, 2026: The gpt-5.2-chat-latest and gpt-5.3-chat-latest snapshots will be removed, with gpt-5.5 as the direct replacement .
• October 23, 2026: The o3-mini-2025-01-31 model and other legacy models like gpt-3.5-turbo-0125 will reach end of life, with gpt-5.5 as the recommended migration path .

The pattern is clear: the API is on a longer, more methodical, staggered schedule, while the consumer-facing ChatGPT has been used as the sharp point of the spear to push users into the latest generation of models as quickly as possible .