Announced in 2016, Gym is an open-source Python library created to help with the advancement of reinforcement knowing algorithms. It aimed to standardize how environments are defined in AI research study, making research more quickly reproducible [24] [144] while providing users with an easy interface for interacting with these environments. In 2022, new developments of Gym have actually been moved to the library Gymnasium. [145] [146]
Gym Retro

Released in 2018, Gym Retro is a platform for reinforcement learning (RL) research study on video games [147] utilizing RL algorithms and study generalization. Prior RL research study focused mainly on enhancing agents to resolve single tasks. Gym Retro gives the capability to generalize between video games with similar concepts but various looks.

RoboSumo

Released in 2017, RoboSumo is a virtual world where humanoid metalearning robotic representatives at first do not have knowledge of how to even walk, but are given the goals of finding out to move and to press the opposing representative out of the ring. [148] Through this adversarial knowing process, the agents find out how to adjust to changing conditions. When an agent is then removed from this virtual environment and placed in a brand-new virtual environment with high winds, the representative braces to remain upright, suggesting it had actually discovered how to balance in a generalized method. [148] [149] OpenAI’s Igor Mordatch argued that competition in between agents could develop an intelligence “arms race” that might increase an agent’s capability to function even outside the context of the competitors. [148]
OpenAI 5

OpenAI Five is a group of 5 OpenAI-curated bots used in the competitive five-on-five computer game Dota 2, that find out to play against human players at a high ability level completely through experimental algorithms. Before ending up being a group of 5, the very first public demonstration took place at The International 2017, the annual best champion competition for the video game, where Dendi, a professional Ukrainian player, lost against a bot in a live individually matchup. [150] [151] After the match, CTO Greg Brockman explained that the bot had discovered by playing against itself for 2 weeks of actual time, which the learning software was an action in the instructions of developing software application that can deal with intricate jobs like a cosmetic surgeon. [152] [153] The system utilizes a form of reinforcement knowing, as the bots find out in time by playing against themselves hundreds of times a day for months, and are rewarded for actions such as eliminating an opponent and taking map objectives. [154] [155] [156]
By June 2018, the ability of the bots expanded to play together as a full group of 5, and they had the ability to defeat groups of amateur and semi-professional gamers. [157] [154] [158] [159] At The International 2018, OpenAI Five played in two exhibition matches against professional players, but ended up losing both video games. [160] [161] [162] In April 2019, OpenAI Five defeated OG, the ruling world champions of the game at the time, 2:0 in a live exhibit match in San Francisco. [163] [164] The bots’ final public appearance came later that month, where they played in 42,729 overall video games in a four-day open online competition, winning 99.4% of those video games. [165]
OpenAI 5’s systems in Dota 2’s bot player reveals the challenges of AI systems in multiplayer online fight arena (MOBA) video games and how OpenAI Five has actually demonstrated the use of deep support knowing (DRL) agents to attain superhuman proficiency in Dota 2 matches. [166]
Dactyl

Developed in 2018, Dactyl uses machine discovering to train a Shadow Hand, a human-like robot hand, to manipulate physical objects. [167] It discovers completely in simulation using the exact same RL algorithms and training code as OpenAI Five. OpenAI tackled the things orientation issue by utilizing domain randomization, a simulation approach which exposes the learner to a variety of experiences rather than trying to fit to reality. The set-up for Dactyl, aside from having movement tracking cams, also has RGB cameras to permit the robot to control an approximate item by seeing it. In 2018, OpenAI showed that the system was able to control a cube and an octagonal prism. [168]
In 2019, OpenAI showed that Dactyl might solve a Rubik’s Cube. The robot had the ability to resolve the puzzle 60% of the time. Objects like the Rubik’s Cube present complex physics that is harder to design. OpenAI did this by enhancing the toughness of Dactyl to perturbations by utilizing Automatic Domain Randomization (ADR), a simulation approach of creating gradually more hard environments. ADR differs from manual domain randomization by not needing a human to define randomization ranges. [169]
API

In June 2020, OpenAI announced a multi-purpose API which it said was “for accessing brand-new AI models developed by OpenAI” to let designers contact it for “any English language AI task”. [170] [171]
Text generation

The company has promoted generative pretrained transformers (GPT). [172]
OpenAI’s original GPT design (“GPT-1”)

The initial paper on generative pre-training of a transformer-based language design was composed by Alec Radford and his coworkers, and published in preprint on OpenAI’s website on June 11, 2018. [173] It showed how a generative model of language might obtain world knowledge and procedure long-range dependencies by pre-training on a varied corpus with long stretches of adjoining text.

GPT-2

Generative Pre-trained Transformer 2 (“GPT-2”) is a not being watched transformer language design and the successor to OpenAI’s initial GPT model (“GPT-1”). GPT-2 was revealed in February 2019, with just limited demonstrative versions at first released to the general public. The complete variation of GPT-2 was not right away released due to issue about prospective misuse, consisting of applications for writing fake news. [174] Some specialists expressed uncertainty that GPT-2 positioned a substantial risk.

In response to GPT-2, the Allen Institute for Artificial Intelligence responded with a tool to identify “neural fake news”. [175] Other scientists, such as Jeremy Howard, cautioned of “the innovation to absolutely fill Twitter, email, and the web up with reasonable-sounding, context-appropriate prose, which would hush all other speech and be difficult to filter”. [176] In November 2019, OpenAI released the complete version of the GPT-2 language design. [177] Several websites host interactive demonstrations of different circumstances of GPT-2 and other transformer models. [178] [179] [180]
GPT-2’s authors argue not being watched language designs to be general-purpose students, shown by GPT-2 attaining modern precision and perplexity on 7 of 8 zero-shot jobs (i.e. the model was not further trained on any task-specific input-output examples).

The corpus it was trained on, called WebText, contains a little 40 gigabytes of text from URLs shared in Reddit submissions with a minimum of 3 upvotes. It avoids certain issues encoding vocabulary with word tokens by utilizing byte pair encoding. This permits representing any string of characters by encoding both specific characters and multiple-character tokens. [181]
GPT-3

First explained in May 2020, Generative Pre-trained [a] Transformer 3 (GPT-3) is a not being watched transformer language design and the successor to GPT-2. [182] [183] [184] OpenAI stated that the full variation of GPT-3 contained 175 billion criteria, [184] 2 orders of magnitude larger than the 1.5 billion [185] in the full variation of GPT-2 (although GPT-3 models with as couple of as 125 million parameters were also trained). [186]
OpenAI mentioned that GPT-3 succeeded at certain “meta-learning” jobs and might generalize the function of a single input-output pair. The GPT-3 release paper offered examples of translation and cross-linguistic transfer knowing in between English and Romanian, and between English and German. [184]
GPT-3 significantly improved benchmark results over GPT-2. OpenAI cautioned that such scaling-up of language designs might be approaching or experiencing the fundamental capability constraints of predictive language models. [187] Pre-training GPT-3 needed numerous thousand petaflop/s-days [b] of compute, compared to tens of petaflop/s-days for the full GPT-2 model. [184] Like its predecessor, [174] the GPT-3 trained design was not instantly launched to the public for concerns of possible abuse, although OpenAI planned to enable gain access to through a paid cloud API after a two-month free personal beta that began in June 2020. [170] [189]
On September 23, 2020, GPT-3 was certified solely to Microsoft. [190] [191]
Codex

Announced in mid-2021, Codex is a descendant of GPT-3 that has furthermore been trained on code from 54 million GitHub repositories, [192] [193] and is the AI powering the code autocompletion tool GitHub Copilot. [193] In August 2021, an API was released in private beta. [194] According to OpenAI, the model can develop working code in over a lots programming languages, the majority of efficiently in Python. [192]
Several issues with glitches, style flaws and security vulnerabilities were pointed out. [195] [196]
GitHub Copilot has been accused of releasing copyrighted code, without any author attribution or license. [197]
OpenAI announced that they would terminate assistance for Codex API on March 23, 2023. [198]
GPT-4

On March 14, 2023, OpenAI revealed the release of Generative Pre-trained Transformer 4 (GPT-4), capable of accepting text or image inputs. [199] They revealed that the updated innovation passed a simulated law school bar test with a rating around the leading 10% of test takers. (By contrast, GPT-3.5 scored around the bottom 10%.) They said that GPT-4 might likewise check out, analyze or generate up to 25,000 words of text, and compose code in all major programming languages. [200]
Observers reported that the version of ChatGPT utilizing GPT-4 was an improvement on the previous GPT-3.5-based version, with the caution that GPT-4 retained a few of the issues with earlier revisions. [201] GPT-4 is likewise capable of taking images as input on ChatGPT. [202] OpenAI has decreased to expose different technical details and stats about GPT-4, such as the precise size of the model. [203]
GPT-4o

On May 13, 2024, OpenAI announced and released GPT-4o, which can process and generate text, images and audio. [204] GPT-4o attained cutting edge outcomes in voice, multilingual, and vision benchmarks, setting new records in audio speech recognition and translation. [205] [206] It scored 88.7% on the Massive Multitask Language Understanding (MMLU) criteria compared to 86.5% by GPT-4. [207]
On July 18, 2024, OpenAI released GPT-4o mini, a smaller version of GPT-4o replacing GPT-3.5 Turbo on the ChatGPT user interface. Its API costs $0.15 per million input tokens and $0.60 per million output tokens, compared to $5 and $15 respectively for GPT-4o. OpenAI expects it to be particularly beneficial for enterprises, start-ups and designers looking for to automate services with AI representatives. [208]
o1

On September 12, 2024, OpenAI launched the o1-preview and o1-mini models, which have actually been designed to take more time to consider their actions, resulting in higher precision. These designs are especially effective in science, coding, and reasoning tasks, and were made available to ChatGPT Plus and Employee. [209] [210] In December 2024, o1-preview was changed by o1. [211]
o3

On December 20, 2024, OpenAI unveiled o3, the follower of the o1 thinking design. OpenAI likewise revealed o3-mini, a lighter and much faster version of OpenAI o3. As of December 21, 2024, this model is not available for public usage. According to OpenAI, they are testing o3 and o3-mini. [212] [213] Until January 10, 2025, safety and security scientists had the opportunity to obtain early access to these designs. [214] The model is called o3 instead of o2 to prevent confusion with telecommunications services provider O2. [215]
Deep research

Deep research is a representative developed by OpenAI, revealed on February 2, 2025. It leverages the capabilities of OpenAI’s o3 model to perform substantial web surfing, data analysis, and synthesis, providing detailed reports within a timeframe of 5 to thirty minutes. [216] With searching and Python tools made it possible for, it reached a precision of 26.6 percent on HLE (Humanity’s Last Exam) criteria. [120]
Image classification

CLIP

Revealed in 2021, CLIP (Contrastive Language-Image Pre-training) is a design that is trained to examine the semantic resemblance in between text and images. It can notably be used for image classification. [217]
Text-to-image

DALL-E

Revealed in 2021, DALL-E is a Transformer model that creates images from textual descriptions. [218] DALL-E utilizes a 12-billion-parameter version of GPT-3 to interpret natural language inputs (such as “a green leather bag formed like a pentagon” or “an isometric view of an unfortunate capybara”) and generate matching images. It can create images of realistic items (“a stained-glass window with an image of a blue strawberry”) as well as items that do not exist in truth (“a cube with the texture of a porcupine”). As of March 2021, no API or code is available.

DALL-E 2

In April 2022, OpenAI announced DALL-E 2, an updated variation of the model with more realistic results. [219] In December 2022, OpenAI released on GitHub software application for Point-E, a brand-new fundamental system for transforming a text description into a 3-dimensional design. [220]
DALL-E 3

In September 2023, OpenAI revealed DALL-E 3, a more powerful model better able to create images from complicated descriptions without manual prompt engineering and render complex details like hands and text. [221] It was released to the general public as a ChatGPT Plus function in October. [222]
Text-to-video

Sora

Sora is a text-to-video model that can produce videos based upon short detailed prompts [223] along with extend existing videos forwards or backwards in time. [224] It can generate videos with resolution as much as 1920x1080 or 1080x1920. The optimum length of produced videos is unidentified.

Sora’s advancement group called it after the Japanese word for “sky”, to signify its “limitless innovative capacity”. [223] Sora’s innovation is an adaptation of the technology behind the DALL · E 3 text-to-image model. [225] OpenAI trained the system using publicly-available videos in addition to copyrighted videos licensed for that purpose, but did not expose the number or the precise sources of the videos. [223]
OpenAI showed some Sora-created high-definition videos to the general public on February 15, 2024, stating that it might produce videos approximately one minute long. It also shared a technical report highlighting the techniques utilized to train the design, and oeclub.org the model’s abilities. [225] It acknowledged a few of its drawbacks, including battles mimicing complex physics. [226] Will Douglas Heaven of the MIT Technology Review called the presentation videos “remarkable”, but noted that they need to have been cherry-picked and may not represent Sora’s common output. [225]
Despite uncertainty from some academic leaders following Sora’s public demonstration, noteworthy entertainment-industry figures have revealed significant interest in the technology’s potential. In an interview, actor/filmmaker Tyler Perry revealed his astonishment at the innovation’s capability to produce realistic video from text descriptions, mentioning its potential to reinvent storytelling and material creation. He said that his enjoyment about Sora’s possibilities was so strong that he had chosen to pause strategies for broadening his Atlanta-based motion picture studio. [227]
Speech-to-text

Whisper

Released in 2022, Whisper is a general-purpose speech acknowledgment model. [228] It is trained on a big dataset of diverse audio and is likewise a multi-task design that can carry out multilingual speech recognition along with speech translation and language identification. [229]
Music generation

MuseNet

Released in 2019, MuseNet is a deep neural net trained to anticipate subsequent musical notes in MIDI music files. It can generate songs with 10 instruments in 15 designs. According to The Verge, a tune produced by MuseNet tends to begin fairly but then fall into chaos the longer it plays. [230] [231] In popular culture, initial applications of this tool were utilized as early as 2020 for the web mental thriller Ben Drowned to create music for the titular character. [232] [233]
Jukebox

Released in 2020, Jukebox is an open-sourced algorithm to create music with vocals. After training on 1.2 million samples, the system accepts a category, artist, and a bit of lyrics and outputs tune samples. OpenAI stated the tunes “show regional musical coherence [and] follow traditional chord patterns” however acknowledged that the songs do not have “familiar bigger musical structures such as choruses that repeat” which “there is a substantial gap” between Jukebox and human-generated music. The Verge mentioned “It’s highly impressive, even if the outcomes seem like mushy versions of songs that might feel familiar”, while Business Insider stated “surprisingly, some of the resulting songs are catchy and sound genuine”. [234] [235] [236]
Interface

Debate Game

In 2018, OpenAI launched the Debate Game, which teaches makers to discuss toy problems in front of a human judge. The purpose is to research study whether such a technique may assist in auditing AI decisions and in developing explainable AI. [237] [238]
Microscope

Released in 2020, Microscope [239] is a collection of visualizations of every substantial layer and neuron of 8 neural network designs which are typically studied in interpretability. [240] Microscope was produced to evaluate the features that form inside these neural networks quickly. The models consisted of are AlexNet, VGG-19, various versions of Inception, and different variations of CLIP Resnet. [241]
ChatGPT

Launched in November 2022, ChatGPT is an artificial intelligence tool constructed on top of GPT-3 that supplies a conversational interface that allows users to ask questions in natural language. The system then responds with a response within seconds.