OpenAI has made another round of changes to its model lineup and API pricing, with the most visible update aimed at developers building on GPT-3.5 Turbo. The move lowers the cost of using one of the company’s most widely used models while also updating GPT-4 Turbo to handle some tasks more completely.
Lower API Prices For GPT-3.5 Turbo
The biggest practical change is a price cut for GPT-3.5 Turbo API access. Input prices are dropping by 50%, while output prices are dropping by 25%.
That brings the cost to $0.0005 per thousand tokens in and $0.0015 per thousand tokens out. For developers, those numbers matter because token usage can grow quickly in products that process large amounts of text.
GPT-3.5 Turbo is described as the model most people interact with, usually through ChatGPT. It has also become a popular API choice because it is lower cost and faster than GPT-4 on many tasks.
The price cut is especially relevant for text-heavy applications. When developers use APIs to analyze full papers, books, or other long documents, the amount of input and output text can become a meaningful part of operating cost.
Why The Cut Matters
The change comes as OpenAI continues to adjust both its models and its pricing. In the source article, the lower price is framed not just as a benefit for developers, but also as a signal for where future consumer options may move.
There is also a competitive angle. As open source or self-managed models improve, OpenAI has an incentive to keep customers from leaving. Lower API prices can make it easier for developers to stay with OpenAI’s platform when comparing cost, speed, and quality.
The article also notes another likely factor: better infrastructure and model streamlining. In plain terms, if a company can run models more efficiently, it has more room to reduce prices while continuing to serve the same kinds of workloads.
For teams building with AI, the useful takeaway is simple: GPT-3.5 Turbo is becoming cheaper to use, and that can change the economics of features that were previously too expensive at scale.
New GPT-3.5 Turbo And GPT-4 Turbo Versions
GPT-3.5 Turbo is also getting a new model version called 0125. OpenAI said it includes “various improvements,” though the source article notes that the company did not highlight much more detail about what changed.
The prior version was 0613, so the 0125 update represents a newer checkpoint for developers who track model versions closely. For many teams, versioning matters because model behavior can affect application quality, testing, and reliability.
GPT-4 Turbo is also receiving a new preview model for API use, again labeled 0125. The notable issue addressed here is task completion, especially in areas such as code generation.
This model completes tasks like code generation more thoroughly than the previous preview model and is intended to reduce cases of “laziness” where the model doesn’t complete a task.
That fix is significant because developers rely on models not only for short answers, but also for multi-step outputs. If a model stops early or avoids finishing a requested task, the user experience suffers and developers may need to add extra prompting or retries.
The updated GPT-4 Turbo model remains in preview mode. Separately, GPT-4 Turbo with vision, also referred to as GPT-4 V, is expected to launch to general availability “in the coming months.”
More Tools Around The Core Models
OpenAI also released new and improved text embedding models. These are more relevant to technical users than everyday ChatGPT users, because embeddings are often used behind the scenes in systems that compare, search, or organize text.
The company also released a new version of its free moderation API, which identifies potentially harmful text. Developers experimenting with moderation support are told to look for version 007.
Taken together, the updates show OpenAI working across several layers of its developer platform:
- Lower pricing for GPT-3.5 Turbo API usage.
- A new GPT-3.5 Turbo model version, 0125.
- A GPT-4 Turbo preview update intended to reduce incomplete task handling.
- New and improved text embedding models.
- A refreshed moderation API version, 007.
The headline change is cost, but the broader pattern is platform maintenance. OpenAI is not only changing prices; it is also refining model behavior, updating technical building blocks, and keeping moderation tooling current.
What Developers Should Watch
For developers already using GPT-3.5 Turbo, the new pricing directly affects the cost of sending and receiving tokens through the API. Products that depend on large volumes of text may see the most noticeable impact.
For teams testing GPT-4 Turbo, the 0125 preview model is worth watching because it targets a concrete behavior problem: incomplete work. The source article specifically points to code generation as an example of a task the model should now complete more thoroughly.
The updates do not remove the need for testing. New model versions can bring improvements, but applications that depend on consistent outputs still need to evaluate whether the updated behavior fits their use case.
Even so, the direction is clear from the source: OpenAI is making GPT-3.5 Turbo cheaper, adjusting GPT-4 Turbo’s task completion, and expanding the surrounding API toolkit for developers building text-heavy AI products.