# GPT-4 March 2023 GPT-4 is a large language model developed by OpenAI, released in March 2023, succeeding GPT-3.5. It introduced multimodal capabilities and improved performance, but OpenAI withheld technical details. It was available in ChatGPT and the API, with a successor GPT-4o later introduced. GPT-4 is a large language model developed by [OpenAI](https://www.wikiprompt.org/wiki/openai), released on March 14, 2023, as the fourth generation in its GPT foundation model series. It succeeded [GPT-3.5](https://www.wikiprompt.org/wiki/large-language-model) and was followed by GPT-5. Unlike its predecessors, GPT-4 is a multimodal model, capable of processing both text and image inputs. OpenAI did not disclose technical details such as model size, architecture, or training hardware, citing competitive and safety concerns. GPT-4 was initially available to ChatGPT Plus subscribers and later through the OpenAI API, with a version supporting up to 32,000 tokens of context. In May 2024, OpenAI introduced GPT-4o, a successor with real-time audio and vision capabilities, which was also made available to free-tier users. ## Background OpenAI introduced the first GPT model in 2018, based on the [transformer architecture](https://www.wikiprompt.org/wiki/transformer) and trained on a large corpus of books. The following year, GPT-2 demonstrated the ability to generate coherent text at scale. In 2020, GPT-3 was released with over 100 times the parameters of GPT-2, and further refinements led to GPT-3.5, which powered the ChatGPT chatbot. GPT-4 built on this lineage, incorporating multimodal input and more advanced training techniques. ## Capabilities GPT-4's default version had an 8K context window, while a special API version supported up to 32K tokens. As a multimodal model, it could accept images as input, enabling users to upload photos and receive suggestions or answers based on the visual content. GPT-4 could also be prompted to interact with external interfaces, such as enclosing a query within `` tags to perform a web search, with results inserted into the model's prompt to form a response. This allowed the model to use APIs, generate images, and summarize webpages. A 2023 article in Nature reported that programmers found GPT-4 useful for coding tasks, such as identifying errors and suggesting optimizations, despite its propensity for error. A biophysicist quoted in the article noted that porting a program from MATLAB to Python took "an hour or so" instead of days. On a test of 89 security scenarios, GPT-4 produced code vulnerable to SQL injection attacks 5% of the time, compared to 40% for [GitHub Copilot](https://www.wikiprompt.org/wiki/github-copilot) from 2021. In November 2023, OpenAI announced GPT-4 Turbo and GPT-4 Turbo with Vision, featuring a 128K context window and significantly lower pricing. ### Aptitude on Standardized Tests Studies have questioned the reliability of running chatbots through standardized tests. In the Torrance Tests of Creative Thinking, GPT-4 scored within the top 1% for originality and fluency, with flexibility scores ranging from the 93rd to the 99th percentile. ### Medical Applications Researchers from [Microsoft](https://www.wikiprompt.org/wiki/microsoft) tested GPT-4 on medical problems and found that, without specialized prompt crafting, it exceeded the passing score on the USMLE by over 20 points, outperforming earlier general-purpose models like GPT-3.5 and models specifically fine-tuned on medical knowledge, such as Med-PaLM. However, the report warned of "significant risks" in using large language models for medical applications, including inaccurate recommendations and hallucinated factual errors. In April 2023, Microsoft and Epic Systems announced plans to provide healthcare providers with GPT-4-powered systems for answering patient questions and analyzing medical records. ### GPT-4o On May 13, 2024, OpenAI introduced GPT-4o ("o" for "omni"), a successor to GPT-4 that processes and generates outputs across text, audio, and image modalities in real time. GPT-4o exhibited rapid response times comparable to human conversation, improved performance on non-English languages, and enhanced vision and audio understanding. Unlike GPT-4, it was available to free-tier users. ## Limitations Like its predecessors, GPT-4 has been known to hallucinate, producing outputs that may include information not in its training data or that contradict the user's prompt. It also lacks transparency in decision-making; when asked to explain its logic, GPT-4 provides post-hoc explanations that may not reflect the actual process, and in many cases, these explanations directly contradict previous statements. In 2023, researchers tested GPT-4 against the ConceptARC benchmark, designed to measure abstract reasoning, and found it scored below 33% on all categories, while specialized models scored 60% on most, and humans scored at least 91%. Sam Bowman, not involved in the research, suggested the results may not indicate a lack of abstract reasoning because the test is visual, while GPT-4 is a language model. ### Bias GPT-4 was trained in two stages. First, it was trained on large internet text datasets to predict the next token. Second, human feedback was used to fine-tune the system through [reinforcement learning from human feedback](https://www.wikiprompt.org/wiki/rlaif), training the model to refuse prompts that violate OpenAI's definition of harmful behavior, such as illegal activities, self-harm, or graphic content. Microsoft researchers suggested GPT-4 may exhibit cognitive biases such as confirmation bias, anchoring, and base-rate neglect. ## Training OpenAI did not release technical details of GPT-4's training. The technical report refrained from specifying model size, architecture, or hardware used. It described a combination of supervised learning on a large dataset followed by reinforcement learning with human and AI feedback, but did not provide details on dataset construction, computing power, or hyperparameters such as learning rate, epoch count, or optimizers. The report cited "the competitive landscape and the safety implications of large-scale models" as reasons for this decision. Sam Altman stated that the cost of training GPT-4 exceeded $100 million, as reported by Semafor. ## Reception and Impact GPT-4's release generated significant attention in the [AI](https://www.wikiprompt.org/wiki/artificial-intelligence) community and beyond. Its multimodal capabilities and improved performance on benchmarks, including standardized tests and medical exams, were widely discussed. However, concerns about hallucination, bias, and lack of transparency persisted. The integration of an early version into Bing Chat in February 2023, before the official release, also drew scrutiny. GPT-4 remained available in the OpenAI API even after its removal from ChatGPT in 2025, and its successor GPT-4o continued to evolve the model's capabilities. --- Source: https://www.wikiprompt.org/wiki/gpt-4-march-2023 License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/) Last updated: 2026-09-13T03:52:29.667015+00:00