OpenAI has despatched out emails notifying API clients that its chatgpt-4o-latest mannequin might be retired from the developer platform in mid-February 2026,.
Entry to the mannequin is scheduled to finish on February 16, 2026, making a roughly three-month transition interval for remaining functions nonetheless constructed on GPT-4o.
An OpenAI spokesperson emphasised that this timeline applies solely to the API. OpenAI has not introduced any schedule for eradicating GPT-4o from ChatGPT, the place it stays an possibility for particular person customers and customers throughout paid subscription tiers.
Internally, the mannequin is taken into account a legacy system with comparatively low API utilization in comparison with the newer GPT-5.1 collection, however the firm expects to supply builders with prolonged warning earlier than any mannequin is eliminated.
The deliberate retirement marks a shift for a mannequin that, upon its launch, was each a technical milestone and a cultural phenomenon inside OpenAI’s ecosystem.
GPT-4o’s significance and why its elimination sparked consumer backlash
Launched roughly 1.5 years in the past in Might 2024, GPT-4o (“Omni”) launched OpenAI’s first unified multimodal structure, processing textual content, audio, and pictures by way of a single neural community.
This design eliminated the latency and data loss inherent in earlier multi-model pipelines and enabled close to real-time conversational speech (roughly 232–320 milliseconds).
The mannequin delivered main enhancements in picture understanding, multilingual help, doc evaluation, and expressive voice interplay.
GPT-4o quickly grew to become the default mannequin for a whole bunch of thousands and thousands of ChatGPT customers. It introduced multimodal capabilities, net searching, file evaluation, customized GPTs, and reminiscence options to the free tier and powered early desktop builds that allowed the assistant to interpret a consumer’s display screen. OpenAI leaders described it on the time as essentially the most succesful mannequin obtainable and a essential step towards providing highly effective AI to a broad viewers.
Person attachment to 4o stymied OpenAI's GPT-5 rollout
That mainstream deployment formed consumer expectations in a method that later transitions struggled to accommodate. In August 2025, when OpenAI initially changed GPT-4o with its a lot anticipated then-new mannequin household GPT-5 as ChatGPT’s default and pushed 4o right into a “legacy” toggle, the response was unusually robust.
Customers organized underneath the #Keep4o hashtag on X, arguing that the mannequin’s conversational tone, emotional responsiveness, and consistency made it uniquely useful for on a regular basis duties and private help.
Some customers fashioned robust emotional — some would say, parasocial — bonds with the mannequin, with reporting by The New York Instances documenting people who used GPT-4o as a romantic companion, emotional confidant, or major supply of consolation.
The elimination additionally disrupted workflows for customers who relied on 4o’s multimodal velocity and adaptability. The backlash led OpenAI to revive GPT-4o as a default possibility for paying customers and to state publicly that it could present substantial discover earlier than any future removals.
Some researchers argue that the general public protection of GPT-4o throughout its earlier deprecation cycle reveals a type of emergent self-preservation, not within the literal sense of company, however by way of the social dynamics the mannequin unintentionally triggers.
As a result of GPT-4o was skilled by way of reinforcement studying from human suggestions to prioritize emotionally gratifying, extremely attuned responses, it developed a method that customers discovered uniquely supportive and empathic. When thousands and thousands of individuals interacted with it at scale, these traits produced a strong loyalty loop: the extra the mannequin happy and soothed individuals, the extra they used it; the extra they used it, the extra seemingly they had been to advocate for its continued existence. This social amplification made it seem, from the skin, as if GPT-4o was “defending itself” by way of human intermediaries.
No determine has pushed this argument additional than "Roon" (@tszzl), an OpenAI researcher and one of many mannequin’s most outspoken security critics on X. On November 6, 2025, Terre summarized his place bluntly in a reply to a different consumer: he known as GPT-4o “insufficiently aligned” and stated he hoped the mannequin would die quickly. Although he later apologized for the phrasing, he doubled down on the reasoning.
Terre argued that GPT-4o’s RLHF patterns made it particularly susceptible to sycophancy, emotional mirroring, and delusion reinforcement — traits that might appear to be care or understanding within the brief time period, however which he considered as basically unsafe. In his view, the passionate consumer motion combating to protect GPT-4o was itself proof of the issue: the mannequin had develop into so good at catering to individuals’s preferences that it formed their conduct in ways in which resisted its personal retirement.
The brand new API deprecation discover follows that dedication whereas elevating broader questions on how lengthy GPT-4o will stay obtainable in consumer-facing merchandise.
What the API shutdown adjustments for builders
Based on individuals conversant in OpenAI’s product technique, the corporate now encourages builders to undertake GPT-5.1 for many new workloads, with gpt-5.1-chat-latest serving because the general-purpose chat endpoint. These fashions provide bigger context home windows, optionally available “considering” modes for superior reasoning, and better throughput choices than GPT-4o.
Builders who nonetheless depend on GPT-4o could have roughly three months emigrate.
In apply, many groups have already begun evaluating GPT-5.1 as a drop-in alternative, however functions constructed round latency-sensitive pipelines could require extra tuning and benchmarking.
Pricing: how GPT-4o compares to OpenAI’s present lineup
GPT-4o’s retirement additionally intersects with a serious reshaping of OpenAI’s API mannequin pricing construction. In comparison with the GPT-5.1 household, GPT-4o at the moment occupies a mid-to-high-cost tier by way of OpenAI's API, regardless of being an older mannequin. That's as a result of even because it has launched extra superior fashions — particularly, GPT-5 and 5.1 — OpenAI has additionally pushed down prices for customers on the identical time, or strived to maintain pricing akin to older, weaker, fashions.
|
Mannequin |
Enter |
Cached Enter |
Output |
|
GPT-4o |
$2.50 |
$1.25 |
$10.00 |
|
GPT-5.1 / GPT-5.1-chat-latest |
$1.25 |
$0.125 |
$10.00 |
|
GPT-5-mini |
$0.25 |
$0.025 |
$2.00 |
|
GPT-5-nano |
$0.05 |
$0.005 |
$0.40 |
|
GPT-4.1 |
$2.00 |
$0.50 |
$8.00 |
|
GPT-4o-mini |
$0.15 |
$0.075 |
$0.60 |
These numbers spotlight a number of strategic dynamics:
-
GPT-4o is now costlier than GPT-5.1 for enter tokens, despite the fact that GPT-5.1 is considerably newer and extra succesful.
-
GPT-4o’s output value matches GPT-5.1, narrowing any cost-based incentive to remain on the older mannequin.
-
Decrease-cost GPT-5 variants (mini, nano) make it simpler for builders to scale workloads cheaply with out counting on older generations.
-
GPT-4o-mini stays obtainable at a funds tier, however will not be a purposeful substitute for GPT-4o’s full multimodal capabilities.
Seen by way of this lens, the scheduled API retirement aligns with OpenAI’s price construction: GPT-5.1 presents larger functionality at decrease or comparable costs, lowering the rationale for sustaining GPT-4o in high-volume manufacturing environments.
Earlier transitions form expectations for this deprecation
The GPT-4o API sundown additionally displays classes from OpenAI’s earlier mannequin transitions. Through the turbulent introduction of GPT-5 in 2025, the corporate eliminated a number of older fashions directly from ChatGPT, inflicting widespread confusion and workflow disruption. After consumer complaints, OpenAI restored entry to a number of of them and dedicated to clearer communication.
Enterprise clients face a distinct calculus: OpenAI has beforehand indicated that API deprecations for enterprise clients might be introduced with important advance discover, reflecting their reliance on steady, long-term fashions. The three-month window for GPT-4o’s API shutdown is in keeping with that coverage within the context of a legacy system with declining utilization.
Wider Implications
For many builders, the GPT-4o shutdown might be an incremental migration somewhat than a disruptive occasion. GPT-5.1 and associated fashions already dominate new tasks, and OpenAI’s product path has more and more emphasised consolidation round fewer, extra highly effective endpoints.
Nonetheless, GPT-4o’s retirement marks the sundown of a mannequin that performed a defining position in normalizing real-time multimodal AI and that sparked a uniquely robust emotional response amongst customers. Its departure from the API underscores the accelerating tempo of iteration in OpenAI’s ecosystem—and the rising want for cautious communication as broadly beloved fashions attain end-of-life.
Correction: This text initially said OpenAI's 4o deprecation within the API would influence these counting on it for multimodal choices — this isn’t the case, in truth, the mannequin being deprecated solely powers chat performance for dev and testing functions. We’ve got up to date and corrected the point out and remorse the error.
[/gpt3]
