Leading AI research company providing cutting-edge language models, image generation, speech recognition, and multimodal AI capabilities.

Compared to the second-generation ada embedding model, the performance has been improved.

The DALL·E model, released in November 2022. DALL·E 2 generates images that are more realistic and accurate than the original model, with a resolution increase of four times.

Whisper is a versatile speech recognition model. It has been trained on a large and diverse audio dataset and is a multi-task model capable of performing multilingual speech recognition, speech translation, and language identification.

The latest multimodal content moderation model capable of analyzing text and images.

The latest text-to-speech model optimized for speed.

The latest text-to-speech model optimized for improved quality.

The previous text was limited to content moderation. We expect omni-moderation-* models to become the best default choice for the future.

The latest DALL·E model was launched by OpenAI in November 2023.

The most powerful second-generation embedding model, replacing 16 first-generation models.

OpenAI's latest speech recognition model with multilingual support, significantly improved accuracy and robustness.

GPT-4o mini is OpenAI's newest model afterGPT-4 Omni, supporting both text and image inputs with text outputs.

GPT-4o ("o" for "omni") is OpenAI's latest AI model, supporting both text and image inputs with text outputs.

A generative model developed by OpenAI that generates 3D objects conditioned on text, capable of directly generating parameters of implicit functions that can be rendered as textured meshes and neural radiance fields.

A faster and cheaper version of o1, particularly adept at coding, mathematical, and scientific tasks that do not require extensive general knowledge.

Our o1 model early preview aims to leverage extensive world knowledge to tackle challenging problems.

OpenAI's most advanced embedding model with 3072 dimensions, achieving 54.9% on MIRACL benchmark with Matryoshka learning for flexible dimension reduction.