Mistral 7B

People can download and run it locally or deploy it on cloud infrastructure for language and code tasks.

Availability

Downloadable; available on Hugging Face

Pricing model

Not established

Deployment

Can be used locally or deployed on AWS, GCP, or Azure

Privacy

Not established

Licence

Apache 2.0

• It has 7 billion parameters. ^[1]
• It outperforms available open models up to 13 billion parameters on standard English and code benchmarks. ^[1]

  As part of this effort, we’re releasing today [Mistral 7B]( {{< ref "news/announcing-mistral-7b" >}} ), our first 7B-parameter model, which outperforms all currently available open models up to 13B parameters on all standard English and code benchmarks .

• It can be used for summarization, structuration, and question answering. ^[1]

  Yet, it can be used to solve many tasks: summarisation, structuration and question answering to name a few.

• Mistral 7B has 7.3 billion parameters. ^[2]
• Mistral reports that Mistral 7B outperforms Llama 2 13B on all evaluated benchmarks. ^[2]

  Mistral 7B significantly outperforms Llama 2 13B on all metrics, and is on par with Llama 34B (since Llama 2 34B was not released, we report results on Llama 34B).

• Mistral 7B uses grouped-query attention for faster inference. ^[2]

  Uses Grouped-query attention (GQA) for faster inference

• Mistral reports a 2x speed improvement for sequence length 16k with a 4k sliding window. ^[2]

  In practice, changes made to FlashAttention and xFormers yield a 2x speed improvement for sequence length of 16k with a window of 4k.

• The model is available to download and use locally. ^[2]

  Download it and use it anywhere (including locally) with our reference implementation ,

• The model can be deployed on AWS, GCP, or Azure. ^[2]

  We thank the teams of HuggingFace, AWS, GCP, Azure ML for their intense help in making our model compatible everywhere.

• Build and govern workplace agents with North 2 Daily brief 2026-10-05