AI BriefingUpstage's Solar 10.7B Tops Hugging Face Open LLM Leaderboard
Key point
Upstage's Solar 10.7B took first place on the Hugging Face Open LLM Leaderboard.
Details
Solar 10.7B took first place on the Hugging Face Open LLM Leaderboard with an average score of 74.2. Upstage stated that this 10.7B English LLM was created using depth up-scaling based on an open-source 7B model, and was pretrained on more than 3 trillion tokens.
The model was released under the Apache 2.0 license, and both a pre-trained version and an instruction fine-tuned version are provided. Upstage highlighted Solar as an SLM suited for enterprise environments, explaining that running it locally can reduce data leakage and also mitigate AI hallucination.
Comparison scores are as follows.
- GPT-3.5 Turbo: 71.07
- Llama 2: 67.87
- Qwen 72B: 73.6
- Mixtral 8x7B: 72.62
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.