Skip to content
AI Stack Nest
Open weightschatfunction calling

Mistral 7B Instruct v0.3

Small local assistants, with a permissive license

Specifications

Mistral 7B Instruct v0.3 specifications
PropertyValue
MakerMistral AI
FamilyMistral
Parameters7B
Context window32K
LicenseApache 2.0
Commercial useYes
Best forSmall local assistants, with a permissive license
Last verified17 Nov 2026

Memory needed to run Mistral 7B Instruct v0.3

Memory by precision
PrecisionMemory (GB)
Ollama default download5.9 GB

Check your own computer with the will-it-run checker.

What it is

Mistral 7B Instruct v0.3 is a 7 billion parameter instruction-tuned model from Mistral AI. Compared with the previous version it has an extended vocabulary of 32,768 tokens, uses the v3 tokenizer and supports function calling.

Size on disk

The Ollama library lists a download of about 4.4 GB and a 32K context window.

Things to know

The model card says the model has no moderation mechanisms built in, so add your own filtering if you put it in front of users. The license is Apache 2.0.

How to run it

Run ollama run mistral.

About these notes

Sizes, context length and license were read from the Ollama library page and the Hugging Face model card on 2026-11-17. We have not benchmarked this model ourselves, so we give no speed or quality scores.

Guides and projects using Mistral 7B Instruct v0.3

More models

Official source: model card. Figures come from the vendor and our own tests; always confirm on the model card before building on a model.