Mistral 7B
An open-source language model with 7 billion parameters released by Mistral AI. It achieves strong performance relative to its parameter count, supports local deployment and fine-tuning, and represents Mistral's open-source strategy.
Core Facts
Mistral 7B was released in September 2023, has 7.3 billion parameters, and outperformed competing models with parameter counts of up to 13 billion and even 34 billion on multiple benchmarks.
The key technical innovations of Mistral 7B include Sliding Window Attention (SWA) and Grouped Query Attention (GQA).
Timeline (last 90 days)
Mistral 7B, previously released by Mistral, adopted the relatively permissive Apache 2.0 license.
Open-source models with fewer than 7B parameters, such as Llama 3 and Mistral 7B, combined with 4-bit quantization, can run on a single consumer-grade GPU or a high-performance CPU.
Mistral 7B uses the Apache 2.0 license and is widely regarded as closer to true open source in the genuine sense
Mistral AI's Mistral 7B model employs a Sliding Window Attention mechanism, where each token only attends to neighboring tokens within a fixed window.
Mistral AI was founded in Paris in 2023 by former Meta and Google DeepMind researchers, and Mistral 7B, with 7 billion parameters, matches or even surpasses the performance of competitors at the 13 billion parameter level.
7B-13B级别的开源小模型（如Llama 2 7B、Mistral 7B）参数容量有限，对外部提供的上下文质量形成刚性依赖
All Facts (12)
Mistral 7B was released in September 2023, has 7.3 billion parameters, and outperformed competing models with parameter counts of up to 13 billion and even 34 billion on multiple benchmarks.
80%VerifiedThe key technical innovations of Mistral 7B include Sliding Window Attention (SWA) and Grouped Query Attention (GQA).
75%UnverifiedMistral的7B模型最初通过BitTorrent种子链接发布
90%UnverifiedMistral AI的7B模型在发布时性能超越了两倍参数量的Llama 2 13B
85%UnverifiedOpen-source models such as Mistral 7B, Falcon, Qwen, and DeepSeek emerged one after another following LLaMA, forming an ecosystem that rivals closed-source models.
70%UnverifiedMistral 7B uses the Apache 2.0 license and is widely regarded as closer to true open source in the genuine sense
60%UnverifiedMistral 7B, previously released by Mistral, adopted the relatively permissive Apache 2.0 license.
50%UnverifiedOpen-source models with fewer than 7B parameters, such as Llama 3 and Mistral 7B, combined with 4-bit quantization, can run on a single consumer-grade GPU or a high-performance CPU.
50%UnverifiedMistral AI's Mistral 7B model employs a Sliding Window Attention mechanism, where each token only attends to neighboring tokens within a fixed window.
50%UnverifiedMistral AI was founded in Paris in 2023 by former Meta and Google DeepMind researchers, and Mistral 7B, with 7 billion parameters, matches or even surpasses the performance of competitors at the 13 billion parameter level.
50%Unverified7B-13B级别的开源小模型（如Llama 2 7B、Mistral 7B）参数容量有限，对外部提供的上下文质量形成刚性依赖
50%Unverified一台配备 16GB 内存的普通 Mac 借助 GGUF 量化格式可流畅运行 7B 至 13B 参数的模型
50%