GPT-4o System Card

By OpenAI Aaron Hurst, Adam Lerer, Adam P. Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford and 423 more. 2024

4,622 citations recorded by the source corpus.

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

Subfields: Cardiology and Cardiovascular Medicine, Endocrinology, Diabetes and Metabolism

View Paper: https://www.semanticscholar.org/paper/7943ec4a67151a559b25cd34369e661c9a7924c8

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Endocrinology, Diabetes and Metabolism

1. Xiyang Dai

Citations in the corpus

• HealthBench: Evaluating Large Language Models Towards Improved Human Health (2025, neutral): “These were easy to run and useful for research iteration at the time, but are narrow, do not reflect real workflows, and have become saturated (Hurst et al., 2024; Saab et al., 2024).”
• Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models (2025, neutral): “Recently, emerging application paradigms such as Retrieval-Augmented Generation (RAG) and agent systems, driven by the advancement of large language models (e.g., Qwen3 (Yang et al., 2025), GPT-4o (Hurst et al., 2024)), have introduced new requirements and challenges for text embedding and reranking”
• Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs (2025, positive): “Building on the success of the Phi family of language models, we extend their capabilities to handle additional modalities — such as vision and audio, achieving significant progress akin to private models like GPT [HLG + 24], Claude [Ant24], and Gemini [TGL + 24].”
• SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild (2025, positive): “As described in §2.1, we use GPT-4o to track shifts in reasoning behaviors on OlympiadBench.”
• Structured 3D Latents for Scalable and Versatile 3D Generation (2024, positive): “Moreover, we employed state-of-the-art multimodal model, GPT4o [1], to caption each 3D asset, ensuring precise and detailed text descriptions.”
• InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency (2025, neutral): “This performance even surpasses GPT-4o [97] (15.18), highlighting that InternVL3.”
• CosyVoice 2: Scalable Streaming Speech Synthesis with Large Language Models (2024, neutral): “The first-package latency is an important metric for streaming synthesis models, which significantly affects the user experience especially in LLM-based voice chat applications, such as GPT-4o [36].”
• Video-R1: Reinforcing Video Reasoning in MLLMs (2025, critical): “Notably, Video-R1-7B attains 35.8% accuracy on VSI-Bench, a challenging video spatial reasoning benchmark, outperforming even proprietary models like GPT-4o [9].”
• Flow-GRPO: Training Flow Matching Models via Online RL (2025, neutral): “M) [4] from 63% to 95% on GenEval, outperforming the state-of-the-art GPT-4o [18] model.”
• R1-VL: Learning to Reason with Multimodal Large Language Models via Step-Wise Group Relative Policy Optimization (2025, neutral): “Multimodal large language models (MLLMs) have achieved significant progress in vision-language understanding [1, 7, 13, 16, 19, 34, 40, 46].”
• LLMs Get Lost In Multi-Turn Conversation (2025, neutral): “We selected a total of 15 LLMs from eight model families: OpenAI (GPT-4o-mini, GPT-4o [30], o3 [57], and GPT-4.”
• InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models (2025, neutral): “5-VL [7] and remains on par with closed-source models ( e.g. , ChatGPT-4o [98], Claude-3.”
• LlaVA-CoT: Let Vision Language Models Reason Step-By-Step (2024, positive): “As shown in Figure 3, since no multimodal model currently exists that can directly produce systematic, structured reasoning, we use GPT-4o [3] to generate detailed reasoning processes, including summary, caption, reasoning, and conclusion, and compile these into the LLaVA-CoT-100k dataset, which we…”
• VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning (2025, neutral): “These models engage in extended reasoning and reflection before arriving at a final answer, in contrast to fast-thinking models like GPT-4o [Hurst et al., 2024] and Claude-3.”
• Thinking in Space: How Multimodal Large Language Models See, Remember, and Recall Spaces (2024, neutral): “5 [78] and GPT-4o [34].”
• The Lessons of Developing Process Reward Models in Mathematical Reasoning (2025, neutral): “For the LLM prompted as Critic Models, i.e., LLM-as-a-judge, we compare with proprietary language models GPT-4o-0806 (Hurst et al., 2024) and o1-mini (OpenAI, 2024), open-source language models Llama-3.”
• WebThinker: Empowering Large Reasoning Models with Deep Research Capability (2025, neutral): “3-70B-Instruct [10]) and closed-source models (DeepSeek-R1-671B [5], GPT-4o [20], o1-preview [35], o3-mini [37], Gemini-2.”