The AI landscape is developing rapidly and nele.ai is keeping its finger on the pulse.
We are pleased to add Anthropic's brand new Claude 3 language models - Opus, Sonnet and Haiku - to our offering. Claude3 complements the existing models from OpenAI, Microsoft and other vendors, such as Llama2, Mistral and Falcon, which we use for processing on dedicated servers.
The Claude 3 models from Anthropic are among the most powerful large voice models currently available. In tests conducted by the LMSYS Chatbot Arena, they perform slightly to significantly better than ChatGPT and other models. In particular, the flagship model Claude 3 Opus with 200,000 tokens per chat achieves the top position in the Elo rating with 1255 points.
Claude 3's strengths lie in the quality of its responses, its understanding of complex relationships and its ability to deliver consistent outputs across multiple iterations. Anthropic's focus on ethical training and safety is reflected in the models.
In addition to their impressive performance, the Claude 3 models also impress with their extremely economical resource consumption. Compared to similar models, the Haiku model only consumes around 2% of the credits per conversation. The models were also trained with training data up to August 2023 and are therefore very up-to-date.
Our goal at nele.ai is to always provide you with the most powerful and advanced AI models. The integration of Claude 3 is another important step that will be followed by others in the near future. For example, we are already checking Google's models from the Gemini project for data security.
Test the performance of Claude 3 directly in your nele.ai environment. We are looking forward to your feedback!
Here is a comparison of the models available in nele.ai according to the Elo score (the LMSYS Chatbot Arena is a crowdsourced open platform for rating Large Language Models (LLMs) - over 500,000 human ratings have been collected here to rank LLMs with the Elo rating system, as of March 29, 2024).
Order descending from highest to lowest score
Elo score: 1255
Token volume per chat: 200,000
Credit consumption factor: 1
Training dates until: August 2023
Elo score: 1252
Token volume per chat: 128,000
Credit consumption factor: 0.5
Training dates until: April 2023
Elo score: 1252
Token volume per chat: 128,000
Credit consumption factor: 0.5
Training dates until: April 2023
Elo score: 1200
Token volume per chat: 200,000
Credit consumption factor: 0.2
Training dates until: August 2023
Elo score: 1185
Token volume per chat: 8,192
Credit consumption factor: 1
Training dates until: September 2021
Elo score: 1185
Token volume per chat: 8,162
Credit consumption factor: 1
Training dates until: September 2021
Elo score: 1185
Token volume per chat: 32,768
Credit consumption factor: 2
Training dates until: September 2021
Elo score: 1177
Token volume per chat: 200,000
Credit consumption factor: 0.02
Training dates until: August 2023
Elo score: 1115
Token volume per chat: 4,096
Credit consumption factor: 0.04
Training dates until: September 2021
Elo score: 1115
Token volume per chat: 4,096
Credit consumption factor: 0.04
Training dates until: September 2021
Elo score: 1115
Token volume per chat: 16,385
Credit consumption factor: 0.08
Training dates until: September 2021
The complete evaluation of the LMSYS Chatbot Arena can be found on this website.