Monday, May 27, 2024

Announcing GPT-4o: OpenAI’s new flagship model on Azure AI

Today, ChatGPT is beginning to push out GPT-4o’s text and image capabilities. OpenAI is launching GPT-4o in the free tier and offering up to five times higher message limits to Plus customers. In the upcoming weeks, ChatGPT Plus will launch an early version of a new Voice Mode that integrates GPT-4o.

GPT-4, OpenAI’s newest deep learning scaling milestone. GPT-4 is a large multimodal model that handles image and text inputs and outputs text. While less proficient than humans in many real-world situations, it performs at human levels on professional and academic benchmarks. It scores in the top 10% of simulated bar exam takers, while GPT-3.5 scores in the bottom 10%. After six months of progressively aligning GPT-4 utilising learning from our adversarial testing programme and ChatGPT, OpenAI achieved their best-ever results on factuality, steerability, and guardrail refusal.

Over two years, OpenAI updated their deep learning stack and co-designed a supercomputer with Azure for their workload. For the system’s first “test run,” OpenAI trained GPT-3.5 last year. Some flaws were resolved and their theoretical underpinnings enhanced. Thus, OpenAI’s GPT-4 training run was unprecedentedly steady, becoming OpenAI’s first huge model whose training performance OpenAI could precisely anticipate. As OpenAI focus on dependable scalability, OpenAI want to improve our technique to foresee and plan for future capabilities earlier, which is crucial for safety.

GPT-4 text input is coming to ChatGPT and the API (with a waiting).OpenAI is working with one partner to make picture input available to more people. OpenAI also open-sourcing OpenAI Evals, their platform for automatic AI model performance review, so anyone may report model flaws to help us improve.


With its ability to receive any combination of text, audio, and image as input and produce any combination of text, audio, and image outputs, GPT-4o (o stands for “omni”) is a step towards far more natural human-computer interaction. It has a response time of up to 320 milliseconds on average while responding to audio inputs, which is comparable to a human’s response time(opens in a new window) during a conversation. It is 50% less expensive and significantly faster in the API, and it matches GPT-4 Turbo speed on text in non-English languages while maintaining performance on text in English and code. When compared to other models, it excels particularly at visual and audio understanding.

You could speak with ChatGPT using Voice Mode with average latency of 2.8 seconds (GPT-3.5) and 5.4 seconds (GPT-4) before GPT-4o. Voice Mode does this by using a pipeline made up of three different models: GPT-3.5 or GPT-4 takes in text and outputs text, a third basic model translates that text back to audio, and a simple model transcribes audio to text. The primary source of intelligence, GPT-4, loses a lot of information as a result of this process. It is unable to directly perceive tone, numerous speakers, background noise, or laughter or emotion expression.

By using it, OpenAI were able to train a single new model end-to-end for text, vision, and audio, which means that the same neural network handles all inputs and outputs. Since GPT-4o is their first model to incorporate all of these modalities, OpenAI have only begun to explore the capabilities and constraints of the model.

Evaluations of models

It surpasses previous standards in terms of multilingual, audio, and visual capabilities, while achieving GPT-4 Turbo-level performance in terms of text, reasoning, and coding intelligence.

Tokenization of language

These 20 languages were selected to serve as an example of how the new tokenizer compresses data across various language families.

Gujarati 4.4x fewer tokens (from 145 to 33)

હેલો, મારું નામ જીપીટી-4o છે. હું એક નવા પ્રકારનું ભાષા મોડલ છું. તમને મળીને સારું લાગ્યું!

Telugu 3.5x fewer tokens (from 159 to 45)

నమస్కారము, నా పేరు జీపీటీ-4o. నేను ఒక్క కొత్త రకమైన భాషా మోడల్ ని. మిమ్మల్ని కలిసినందుకు సంతోషం!

Tamil 3.3x fewer tokens (from 116 to 35)

வணக்கம், என் பெயர் ஜிபிடி-4o. நான் ஒரு புதிய வகை மொழி மாடல். உங்களை சந்தித்ததில் மகிழ்ச்சி!

Marathi 2.9x fewer tokens (from 96 to 33)

नमस्कार, माझे नाव जीपीटी-4o आहे| मी एक नवीन प्रकारची भाषा मॉडेल आहे| तुम्हाला भेटून आनंद झाला!

Hindi 2.9x fewer tokens (from 90 to 31)

नमस्ते, मेरा नाम जीपीटी-4o है। मैं एक नए प्रकार का भाषा मॉडल हूँ। आपसे मिलकर अच्छा लगा!

Urdu 2.5x fewer tokens (from 82 to 33)

ہیلو، میرا نام جی پی ٹی-4o ہے۔ میں ایک نئے قسم کا زبان ماڈل ہوں، آپ سے مل کر اچھا لگا!

Arabic 2.0x fewer tokens (from 53 to 26)

مرحبًا، اسمي جي بي تي-4o. أنا نوع جديد من نموذج اللغة، سررت بلقائك!

Persian 1.9x fewer tokens (from 61 to 32)

سلام، اسم من جی پی تی-۴او است. من یک نوع جدیدی از مدل زبانی هستم، از ملاقات شما خوشبختم!

Russian 1.7x fewer tokens (from 39 to 23)

Привет, меня зовут GPT-4o. Я — новая языковая модель, приятно познакомиться!

Korean 1.7x fewer tokens (from 45 to 27)

안녕하세요, 제 이름은 GPT-4o입니다. 저는 새로운 유형의 언어 모델입니다, 만나서 반갑습니다!

Vietnamese 1.5x fewer tokens (from 46 to 30)

Xin chào, tên tôi là GPT-4o. Tôi là một loại mô hình ngôn ngữ mới, rất vui được gặp bạn!

Chinese 1.4x fewer tokens (from 34 to 24)


Japanese 1.4x fewer tokens (from 37 to 26)


Turkish 1.3x fewer tokens (from 39 to 30)

Merhaba, benim adım GPT-4o. Ben yeni bir dil modeli türüyüm, tanıştığımıza memnun oldum!

Italian 1.2x fewer tokens (from 34 to 28)

Ciao, mi chiamo GPT-4o. Sono un nuovo tipo di modello linguistico, è un piacere conoscerti!

German 1.2x fewer tokens (from 34 to 29)

Hallo, mein Name is GPT-4o. Ich bin ein neues KI-Sprachmodell. Es ist schön, dich kennenzulernen.

Spanish 1.1x fewer tokens (from 29 to 26)

Hola, me llamo GPT-4o. Soy un nuevo tipo de modelo de lenguaje, ¡es un placer conocerte!

Portuguese 1.1x fewer tokens (from 30 to 27)

Olá, meu nome é GPT-4o. Sou um novo tipo de modelo de linguagem, é um prazer conhecê-lo!

French 1.1x fewer tokens (from 31 to 28)

Bonjour, je m’appelle GPT-4o. Je suis un nouveau type de modèle de langage, c’est un plaisir de vous rencontrer!

English 1.1x fewer tokens (from 27 to 24)

Hello, my name is GPT-4o. I’m a new type of language model, it’s nice to meet you!

Availability of the model

OpenAI’s most recent endeavour to expand the capabilities of deep learning this time towards usefulness in real-world applications is GPT-4o. Over the past two years, they have put a lot of effort into increasing efficiency at every stack layer. OpenAI are able to provide a GPT-4 level model to a much wider audience as a first fruit of this study. Iteratively, the capabilities of GPT-4o will be released (with enhanced red team access commencing immediately).

The API lets developers use GPT-4o for text and vision. Compared to GPT-4 Turbo, GPT-4o has five times higher rate limitations, is half the price, and is two times faster. In the upcoming weeks, OpenAI intend to make support for GPT-4o’s enhanced audio and video capabilities available via the API to a select number of reliable partners.

OpenAI, known for ChatGPT, has advanced huge language models with GPT-4o. Multimodal processing and response to text, visuals, and audio make it stand out. The salient characteristics of GPT-4o are as follows:

Essential features:

Multimodal: This is GPT-4o‘s most important feature. It is capable of processing and reacting to audio, pictures, and text. Consider giving it an audio clip and asking it to summarise the conversation, or showing it a picture and asking it to compose a poem about it.

Enhanced performance: According to OpenAI, GPT-4o performs better than its predecessors in a number of domains, including text production, audio processing, image identification, and complicated text interpretation.
Limitations and safety:

Focus on safety: By screening training data and putting safety measures in place, OpenAI puts safety first. Additionally, in order to find any potential problems like bias or manipulation, they have carried out risk assessments and external testing.

Restricted distribution: Currently, GPT-4o’s text and image input/output features are accessible via OpenAI’s API. There may be a subsequent release with audio capability.


Particular skills: It’s uncertain how much GPT-4o can really do when it comes to multimodal reasoning or complicated audio problems.

Long-term effects: It’s too soon to say what practical uses and possible downsides GPT-4o may have.

With great pleasure, Microsoft announces the release of OpenAI’s new flagship model, GPT-4o, on Azure AI. This innovative multimodal model raises the bar for conversational and creative AI experiences by combining text, visual, and audio capabilities. GPT-4o is currently available for preview in the Azure OpenAI Service and supports both text and images.

A breakthrough for Azure OpenAI Service’s generative AI

A change in the way AI models engage with multimodal inputs is provided by GPT-4o. Through the seamless integration of text, graphics, and music, GPT-4o offers a more immersive and dynamic user experience.

Highlights of the launch: Quick access and what to anticipate

Customers of Azure OpenAI Service can now, in two US locations, explore the vast potential of GPT-4o via a preview playground in Azure OpenAI Studio. The model’s potential is shown by this first version, which focuses on text and visual inputs, opening the door for additional features like audio and video.

Effectiveness and economy of scale

The GPT-4o is designed with efficiency and speed in mind. Its sophisticated capacity to manage intricate queries with less resources can result in improved performance and cost savings.

Possible applications to investigate using GPT-4o

The implementation of GPT-4o presents a multitude of opportunities for enterprises across diverse industries:

Improved customer service: GPT-4o allows for more dynamic and thorough customer assistance conversations by incorporating various data inputs.

Advanced analytics: Make use of GPT-4o’s capacity to handle and examine various data kinds in order to improve decision-making and unearth more profound insights.

Content innovation: Create interesting and varied content forms that appeal to a wide range of customer tastes by utilising GPT-4o’s generating capabilities.

Future advancements to look forward to: GPT-4o at Microsoft Build 2024

To assist developers in fully realising the potential of generative AI, Azure is excited to provide additional information about GPT-4o and other Azure AI advancements at Microsoft Build 2024.

Utilise Azure OpenAI Service to get started

Take the following actions to start using GPT-4o and Azure OpenAI Service:

  • Check out GPT-4o in the preview version of the Azure OpenAI Service Chat Playground.
  • If you don’t currently have access to Azure OpenAI Services, fill out this form to request access.
  • Find out more about the most recent improvements to the Azure OpenAI Service.
  • Learn about Azure’s responsible AI tooling with Azure AI Content Safety.
Since June 2023, Drakshi has been writing articles of Artificial Intelligence for govindhtech. She was a postgraduate in business administration. She was an enthusiast of Artificial Intelligence.


Please enter your comment!
Please enter your name here

Recent Posts

Popular Post Would you like to receive notifications on latest updates? No Yes