Top Chat Models (Open-Source)
Optimized for extended conversations, this mix utilizes top-performing models in multi-turn interactions. It leverages the LMSys Chatbot Arena rankings for conversations with two or more turns, ensuring high-quality responses in longer dialogues.
API
Example
Models
This mix uses the models below:
meta-llama/Meta-Llama-3.1-405B-Instruct
Provided by multiple sources
mistralai/large-2
Provided by multiple sources
meta-llama/Meta-Llama-3.1-70B-Instruct
Provided by multiple sources
Qwen/Qwen2.5-72B-Instruct
Provided by multiple sources
Readme
Top Chat Models (LMSys)
Optimized for extended conversations, this mix utilizes top-performing models in multi-turn interactions. It leverages the LMSys Chatbot Arena rankings for conversations with two or more turns, ensuring high-quality responses in longer dialogues. The weight is a function of the Elo score adjusted for number of votes and variance. Models with more consistently high votes will be weighed more heavily.
Categories
- 💬 General
- 🏆 Leaderboard
- 🌐 Open-Source
- 🧠 Reasoning
Composition
This mix produces responses from the following models:
| Model | Weight % |
|---|---|
| "meta-llama/Meta-Llama-3.1-405B-Instruct" | 76.37% |
| "mistralai/large-2" | 11.17% |
| "meta-llama/Meta-Llama-3.1-70B-Instruct" | 8.20% |
| "Qwen/Qwen2.5-72B-Instruct" | 4.27% |
Update Frequency: Based on LMSys Source: https://lmarena.ai/