Comparison of the AI models used by nele.ai

Which AI models does nele.ai offer?

nele.ai provides various AI models for generating texts and images as well as for image recognition (vision).

‍

Text-generating AI models at nele.ai (as of August 2024):

‍Server location Europe

Azure GPT-4o (128k ➜ approx. 96,000 words* per chat) - Trained with data up to October 2023

Azure GPT-4o mini (128k ➜ approx. 96,000 words* per chat) - Trained with data up to October 2023

Azure GPT-4 Turbo (128k ➜ approx 96,000 words* per chat) - Trained with data up to December 2023

Azure GPT-3.5 Turbo (16k ➜ approx. 12,000 words* per chat) - Trained with data up to September 2021

Claude 3.5 Sonnet (200k ➜ approx. 150,000 words per chat) - Trained until August 2023

Claude 3 Haiku (200k ➜ approx. 150,000 words per chat) - Trained until August 2023

Mistral Small (32k ➜ approx. 24,000 words per chat) — training time unknown

Mistral Large (32k ➜ approx. 24,000 words per chat) — training time unknown

‍

Server location USA

GPT-4o (128k ➜ approx. 96,000 words* per chat) - Trained with data up to October 2023

GPT-4o mini (128k ➜ approx. 96,000 words* per chat) - Trained with data up to October 2023

GPT-4 Turbo (128k ➜ approx. 96,000 words* per chat) - Trained with data up to December 2023

GPT-3.5 Turbo (16k ➜ approx. 12,000 words* per chat) - Trained with data until September 2021

Claude 3 Opus (200k ➜ approx. 150,000 words per chat) - Trained until August 2023

Claude 3.5 Sonnet (200k ➜ approx. 150,000 words per chat) - Trained until August 2023

Claude 3 Haiku (200k ➜ approx. 150,000 words per chat) - Trained until August 2023

‍

Image-generating AI models at nele.ai (as of August 2024)

Server location Europe

Azure DALL·E3

‍

Server location USA

OpenAI DALL·E3

‍

Vision (image recognition models) at nele.ai (as of August 2024)

Server location Europe

Azure GPT-4o

Azure GPT-4o mini

Claude 3.5 Sonnet

Claude 3 Haiku

‍

Server location USA

GPT-4o

GPT-4o mini

Claude 3 Opus

Claude 3.5 Sonnet

Claude 3 Haiku

‍

IMPORTANT NOTICE

All content generated by AI must be checked again for accuracy before use.

‍

*Generative AI models are limited by a token limit, where a token is usually a word, part of a word, or a punctuation mark. This limit defines how much text the AI can process or generate in a session. If the interaction reaches this maximum number of tokens, the AI cannot capture any further information without overwriting previous content. The most well-known example of this is GPT-3 Turbo from OpenAI, which has an upper limit of 12,000 tokens. Within this framework, the system must select the most relevant information to continue the conversation effectively.