The OpenAI API is supported by a varied range of GPT Models. Each model offering distinct functionalities and available at various pricing levels. Additionally, these models can be tailored through fine-tuning to meet the specific needs of your application.
| Model | Description |
|---|---|
| GPT-6 | Gathering training data |
| GPT-5 | Under training for today (Q1 2024) |
| GPT-4 & GPT-4 Turbo | Models enhancing GPT-3.5 capabilities, adept in interpreting and producing both text and code. |
| GPT-3.5 Turbo | Models advancing upon GPT-3.5, skilled in comprehending and crafting text and code. |
| GPT base | Models focusing on understanding and generating text and code, sans instruction following. |
GPT-6
Gathering training data
GPT-5
Under training for today (Q1 2024)
GPT-4 & GPT-4 Turbo
GPT-4, along with GPT-4 Turbo, represents the latest advancement in OpenAI’s suite of multimodal models. They capable of interpreting both textual and visual inputs to generate textual outputs. These models mark a significant leap over their predecessors by solving complex problems with unmatched accuracy, thanks to their extensive knowledge base and superior reasoning skills. Available to subscribers via the OpenAI API. GPT-4 is particularly designed for chat applications, though it is equally adept at handling traditional completion tasks. Detailed instructions for employing GPT-4 in text generation projects are provided for users’ benefit.
Here’s a breakdown of the various GPT-4 models:
| Model | Description | Context Window | Training Data |
|---|---|---|---|
| gpt-4-0125-preview | The latest iteration of GPT-4 Turbo designed to minimize “laziness,” capable of producing up to 4,096 tokens. | 128,000 tokens | Up to Dec 2023 |
| gpt-4-turbo-preview | Directs to gpt-4-0125-preview. | 128,000 tokens | Up to Dec 2023 |
| gpt-4-1106-preview | GPT-4 Turbo variant with enhanced features like JSON mode, consistent outputs, and parallel processing, offering up to 4,096 tokens. | 128,000 tokens | Up to Apr 2023 |
| gpt-4-vision-preview | GPT-4 model capable of processing images alongside possessing all Turbo features, linking to gpt-4-1106-vision-preview. | 128,000 tokens | Up to Apr 2023 |
| gpt-4-1106-vision-preview | GPT-4 variant with image understanding and full Turbo functionalities, also generating up to 4,096 tokens. | 128,000 tokens | Up to Apr 2023 |
| gpt-4 | Refers to gpt-4-0613, indicating continuous model enhancements. | 8,192 tokens | Up to Sep 2021 |
| gpt-4-0613 | A snapshot of GPT-4 from June 13th, 2023, with improved function calling feature. | 8,192 tokens | Up to Sep 2021 |
| gpt-4-32k | Points to gpt-4-32k-0613, a model superseded by GPT-4 Turbo despite its upgrades. | 32,768 tokens | Up to Sep 2021 |
| gpt-4-32k-0613 | A version of gpt-4-32k from June 13th, 2023, enhanced for better function calling, not widely released. | 32,768 tokens | Up to Sep 2021 |
GPT-4 Multilingual capabilities
GPT-4 surpasses earlier large language models and, as of 2023, most cutting-edge systems that typically rely on training tailored to specific benchmarks or extensive manual configuration. Utilizing the MMLU benchmark—a collection of multiple-choice questions in English spanning 57 topics—GPT-4 not only significantly exceeds the capabilities of previous models in English but also showcases robust proficiency across various other languages. This achievement highlights GPT-4’s versatility and its advanced understanding and processing of multilingual content.
GPT-3.5 Turbo
GPT-3.5 Turbo models are designed to comprehend and produce both natural language and code, having been fine-tuned for chat interactions through the Chat Completions API. Despite this specialization, they are also highly effective in handling tasks beyond chat-based applications.
Here’s the information presented in a structured table format along with a slight rephrasing:
| Model | Description | Context Window | Training Data |
|---|---|---|---|
| gpt-3.5-turbo-0125 | An upgraded version of GPT-3.5 Turbo that enhances accuracy in generating responses in specified formats and addresses a bug affecting text encoding for non-English function calls. Capable of producing up to 4,096 tokens. | 16,385 tokens | Up to Sep 2021 |
| gpt-3.5-turbo | Currently linked to gpt-3.5-turbo-0613. It will automatically update from gpt-3.5-turbo-0613 to gpt-3.5-turbo-0125 on February 16th. | 4,096 tokens | Up to Sep 2021 |
| gpt-3.5-turbo-1106 | This model version of GPT-3.5 Turbo has been refined for better adherence to instructions, JSON mode, consistent outputs, and simultaneous function execution. It also supports up to 4,096 output tokens. | 16,385 tokens | Up to Sep 2021 |
| gpt-3.5-turbo-instruct | Features capabilities akin to GPT-3 era models. It is designed for the legacy Completions endpoint rather than the Chat Completions. | 4,096 tokens | Up to Sep 2021 |
| gpt-3.5-turbo-16k | Legacy model currently referring to gpt-3.5-turbo-16k-0613. | 16,385 tokens | Up to Sep 2021 |
| gpt-3.5-turbo-0613 | Legacy snapshot as of June 13th, 2023, set to be deprecated on June 13, 2024. | 4,096 tokens | Up to Sep 2021 |
| gpt-3.5-turbo-16k-0613 | Legacy snapshot of gpt-3.5-16k-turbo from June 13th, 2023, also due for deprecation on June 13, 2024. | 16,385 tokens | Up to Sep 2021 |
Read more about GPTs in our Blog.