Models
Model Guides
The models that are being used in our platform are listed below in detail.
Model |
Company |
Modality |
Type |
License |
|---|---|---|---|---|
OpenAI |
text |
generative |
Proprietary |
|
OpenAI |
multimodal |
generative |
Proprietary |
|
OpenAI |
multimodal |
generative |
Proprietary |
|
Amazon Bedrock |
text |
generative |
Proprietary |
|
Google VertexAI |
text |
generative |
Proprietary |
|
Google VertexAI |
multimodal |
generative |
Proprietary |
|
Google VertexAI |
multiimodal |
generative |
Proprietary |
|
Cerebras Systems |
text |
generative |
Open-Source |
|
OpenAI |
text |
embedding |
Proprietary |
|
Google VertexAI |
text |
embedding |
Proprietary |
|
OpenAI |
multimodal |
generative |
Proprietary |
|
Stability AI |
multimodal |
generative |
Open-Source |
Model Description
gpt-35-turbo-16k
The model is OpenAI model, pretrained on large dataset. GPT-3.5 Turbo with a 16k token context window is an extended version of GPT-3.5 designed to handle larger inputs and provide more coherent and comprehensive outputs. It’s suitable for more complex conversations and tasks involving substantial amounts of text. The max_tokens is 16384 and max_output_token is 4096.
gpt-4
GPT-4 with a 128k token context length provides a significant advantage in handling extremely large documents or conversations. This model can maintain long-form continuity and contextual accuracy over extended interactions. The max_tokens is 128000 and max_output_token is 4096. The input and output type can be text or image.
gpt-4o
This version of GPT-4, optimized for even larger contexts, supports up to 128k tokens, making it ideal for tasks involving large amounts of data or long-term context maintenance. It is an ideal choice for intricate applications that require substantial context retention. The max_tokens is 128000 and max_output_token is 4096. The input and output type can be text or image.
amazon.titan-tg1-large
Titan-TG1-Large is one of Amazon’s large language models, available via AWS. It is designed for various business and conversational applications, with solid performance for handling relatively large amounts of text but with a more limited context length compared to other models. The max_tokens is 4096 and max_output_token is 2048.
gemini-pro
Gemini Pro is part of Google’s Gemini family of models, aimed at providing high-quality natural language processing with a focus on productivity and creative tasks. It’s optimized for handling long inputs, such as articles or reports. The max_tokens is 32760 and max_output_token is 8192.
gemini-1.5-pro
Gemini 1.5 Pro is an upgraded version of Google’s Gemini series, offering improved performance for professional and creative applications. It supports advanced language understanding with a focus on generating high-quality text for various industries. The max_tokens is 2097152 and max_output_token is 8192. The input type can be text, code, images, audio, video, video with audio, PDF whereas the output type is text only.
gemini-1.5-flash
Gemini 1.5 Flash is a lighter version of the Gemini 1.5 Pro model, optimized for faster responses and efficient handling of shorter inputs. Ideal for applications that require quick, real-time interactions with less need for long-term context. The max_tokens is 1048576 and max_output_token is 8192. The input type can be text, code, images, audio, video, video with audio, video, video with audio or PDF whereas the output type is text only.
jais-30b-chat
Jais 30B Chat is a large language model with 30 billion parameters, designed for conversational AI applications. It has a medium context window suited for various chatbot and interactive applications. The max_tokens is 8192 and max_output_token is 2048.
text-embedding-ada-002
text-embedding-ada-002 is an embedding model developed by OpenAI, designed to convert text into high-dimensional vector representations. These vectors are useful for tasks like semantic search, clustering, classification, and similarity comparisons. It’s an advanced version of the Ada embedding series, with a focus on performance and accuracy in embedding generation. It processes textual input and outputs dense vector representations that can capture the semantic meaning of the input text. The dimension size is 1536.
textembedding-gecko
textembedding-gecko is a specific version of a pre-trained text embedding model offered by Google Cloud Platform (GCP) through Vertex AI and their suite of AI tools and services. The dimension size is 768.
dalle
DALLE is a powerful and creative text-to-image model from OpenAI, known for generating imaginative and realistic images from detailed text prompts. It’s primarily closed-source and available through OpenAI’s API. The input and output type can be images or text.
stable-diffusion
Stable Diffusion, on the other hand, is an open-source text-to-image model that uses a latent diffusion approach to generate high-quality images. It provides flexibility and customization options and has gained traction for being open-source, allowing widespread usage, fine-tuning, and integration in various applications. The input and output type can be images or text