Join our daily and weekly newsletters to obtain the latest updates and exclusive content on the coverage of the industry leader. Get more information
Operai is implementing GPT-4.1, its new non-reasonable large (LLM) model that balances high performance with lower cost, to Chatgpt users. The company is starting with its subscribers who pay in Chatgpt Plus, Pro and Team, with access to the users of the company and the expected education in the coming weeks.
It is also adding GPT-4.1 Mini, which replaces GPT-4O MINI as the default value for all chatgpt users, including free levels. The “Mini” version provides a parameter on a smaller scale and, therefore, a less powerful version with similar safety standards.
The Models Are Both Avarable via the “More Models” Dropdown Selection in the Top Corner of the Chat Window Chatgpt, Giving Users Flexibility to Choose Between GPT-4.1, GPT-4.1 Mini, and Reasoning Modi-High O4-mini, and and and o4i, and and o4i, and and o4i, and and o4i, and and o4i, and and o4i, and and o4i, and o4-mini, and and o4i, and o4i, and o4-mini, and o4i, and o4-mini, and o4i, and o4i, and o4-mini, and o4i, and o4i, and o4i, and O4-mini, and o4-mini, and o4-mini, and o4i, and o4i, and o4i, and o4i, and o4i, and o4i, and o4i, and o4i, and o4i, and o4i.
Initially intended to use only for third-party software and IA developers through the OpenAight Application Programming Interface (API), GPT-4.1 was added to ChatgPT following strong user comments.
The leader of the Operai post Michelle Pokrass training confirmed in X the change was driven by the demand, writing: “Initially we were planning to keep this model of model alone, but they all wanted it in ChatgPT 🙂 Happy codification!”
Openai’s product director Kevin Weil published in X saying: “We built it for developers, so it is very good to encode and instruct following, try it!”
A company -centered model
GPT-4.1 was designed from scratch for business degree practicality.
Released in April 2025 together with GPT-4.1 Mini and Nano, this model family prioritized the needs of developers and production use cases.
GPT-4.1 It offers an improvement of 21.4 points in GPT-4O at the Swe-Bench verified software reference point, and a gain of 10.5 points in the instruction following the tasks in the point of reference of Multichallenge de Scale. It also reduces verms by 50% compared to other models, a feature of the company’s users praised the early tests of the duration.
Context, speed and access to the model
GPT-4.1 Admits the standard Windows context for chatgpt: 8,000 tokens for free users, 32,000 tokens for more and 128,000 tokens for PRO users.
According to developer Angel Bogado publication in X, these limits coincide with those used by previous chatpt models, thought plans are underway to further increase the size of the context.
While GPT-4.1 API versions can process up to one million tokens, this expanded capacity is not yet notable in Chatgpt, he thought that future support has been suggested.
This extended context capacity allows API users to feed whole or large legal and financial documents in the model, or grateful to review multiple document contracts or analyze large record files.
Operai has recognized some performance decrease with extremely large tickets, but business test cases suggest solid performance of up to several hundred thousand tokens.
Evaluations and security
Openais also launched a Safety Evaluations Hub website to provide users access to key performance metrics on all models.
GPT-4.1 shows solid results in thesis evaluations. In the precision tests of fact, he obtained 0.40 at the simpleqa and 0.63 reference point in Personqa, surpassing several predecessors.
He also obtained 0.99 to Openi’s “non -insecure” extent in standard rejection tests, and 0.86 in more challenging indications.
However, in the Jailbreak StrongrJect test, an academic reference point for security under adverse conditions, GPT-4.1 obtained 0.23, behind models such as GPT-4O-MINI and O3.
That said, he obtained a strong 0.96 in Jailbreak indications of human origin, indicating a more robust security of the real world under typical use.
In compliance with the instructions, GPT-4.1 follows the hierarchy defined by OpenAIS (developer system, developer on user messages) with a score or 0.71 to resolve system messages conflicts against user. It also works well to protect protected phrases and avoid gifts of solutions in tutoring scenarios.
GPT-4.1 contextualization against predecessors
The launch of GPT-4.1 occurs after the scrutiny around GPT-4.5, which debuted in February 2025 as a preview of research. That model emphasized better learning without supervision, a richer knowledge base and reduced hallucinations that fell from 61.8% in GPT-4o to 37.1%. He also showed improvements in emotional nuances and writing long, but many users found subtle improvements.
Despite these profits, GPT-4.5 caused criticism for its high price at $ 180 per million production tokens through an Apidanderming yield in mathematics and coding reference points in relation to O-series or OpenAI models. Industry figures indicated that although GPT-4.5 was stronger in the general conversation and content generation, it had a lower performance in specific developer applications.
On the contrary, GPT-4.1 is intended as a faster and more focused alternative. Although it lacks the knowledge amplitude of GPT-4.5 and an extensive emotional modeling, it is better tuned for practical coding assistance and adheres more reliable to the user’s instructions.
In the OpenAI API, GPT-4.1 is currently a price at $ 2.00 per million input tokens, $ 0.50 per million input tokens stored in cache and exit tokens of $ 8.00 per million.
For those looking for a balance between speed and intelligence at a lower cost, GPT-4.1 MINI is available at $ 0.40 per million input tokens, $ 0.10 per million tokens of cache in exit and tokens outkens of $ 1.60 per million.
Google Flash-Lite and Flash models are available at $ 0.075– $ 0.10 per million input tokens and $ 0.30– $ 0.40 per million output tokens, less than a tenth part of the cost of GPT-4.1 base fees.
But while GPT-4.1 has a higher price, it offers stronger software engineering points and a more precise instruction later, which can be critical for business implementation scenarios that require reliability about cost. Ultimately, Openai GPT-4.1 offers a premium experience for precision and development performance, while Google Gemini models attract conscious costs of costs that need levels of flexible models and multimodal capabilities.
What means for business decision makers
The introduction of GPT-4.1 provides specific benefits to business teams that manage the implementation of LLM, orchestration and data operations:
- IA engineers supervising the implementation of LLM You can expect an improved speed and adhesion instruction. For the equipment that manages the LLM LifeCle-from the fine adjustment model to the problem solving-GPT-4.1 a set of more receptive and efficient tools. It is privileged for read -pressure equipment to send high performance models quickly without compromising safety or compliance.
- The orchestration of AI leads Focused on the design of scalable pipes will appreciate the robustness of GPT-4.1 against most user-induced failures and their strong performance in messages of messages. This facilitates integration into orchestration systems that prioritize the consistency, model validation and operational reliability.
- Data engineers Responsible for high quality data caressing and integrating new tools will benefit from the lowest GPT-4.1 hallucination rate and greater objective precision. Its most predictable output behavior helps build reliable data workflows, even when equipment resources are built.
- IT security professionals The task of integrating safety into devotes pipes can find value in the resistance of GPT-4.1 to common Jailbreaks and its controlled output behavior. While its academic jailbreak resistance score leaves space to improve, the high performance of the model against human origin helps support safe integration into internal tools.
In all thesis roles, the positioning of GPT-4.1 as an optimized model for more clarity, compliance and implementation efficiency makes it a convincing option for medium-sized companies that seek to balance performance with operational demands.
A new step forward
While GPT-4.5 represented a milestone in the model development, GPT-4.1 focuses on utility. It is not the most extent or more multimodal, but offers significant profits in areas that matter for companies: precision, implementation efficiency and cost.
This repositioning reflects a broader industry trend, far from building larger models at any cost, and towards the most accessible and adaptable capable models. GPT-4.1 meets that need, offering a flexible tool and ready for production for equipment that attempt to integrate the AI most deeply in its commercial operations.
As Openai continues to evolve its model actions, GPT-4.1 repeats a step forward in the democratization of advanced for business environments. For the balance capacity of decision makers with the ROI, it offers a clearer path towards deployment without sacrificing performance or security.
