Mistral releases 24B open-source model Mistral Small 3, matching Llama 3.3 70B performance under Apache 2.0
Original title:Research Mistral Small 3 January 30, 2025 By Mistral Team
AI summary
Mistral has released Mistral Small 3 with 24B parameters, open-sourcing both pretrained and instruction-tuned weights under the Apache 2.0 license. It says the model achieves over 81% accuracy on MMLU with 150 tokens/s latency, performs comparably to Llama 3.3 70B instruct while being more than 3x faster on the same hardware, and can run locally after quantization on a single RTX 4090 or a Mac with 32GB of memory.
Key Takeaways
- Mistral has released Mistral Small 3 with 24B parameters,open-sourcing both pretrained and…
- 0 license.
- It says the model achieves over 81% accuracy on MMLU with 150 tokens/s latency,performs comparably…
- 3 70B instruct while being more than 3x faster on the same hardware,and can run locally after…
Quick Facts
- Published
- Updated
- Source
- Mistral AI:News(网页)
- Category
- Model Releases
- AI score
- 72
官方完整说明了模型规格、性能数字和开放部署渠道,读者可据此评估本地部署与开源基座的可行性。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Mistral releases 24B open-source model Mistral Small 3, matching Llama 3.3 70B performance under Apache 2.0. January 30, 2025. https://new.nat6.net/en/news/cmppdcr7e0e42slv4ke38ig2c/