Skip to main content
Rase AI Intelligence Newsletter

Models & assistants Model or AI service

Mistral 7B made a compact language model downloadable

Mistral AI released Mistral 7B’s downloadable weights in September 2023 under Apache 2.0. A team could run the model on its own infrastructure or adapt the weights for a particular application. Mistral retired its hosted v0.1 offering in 2025 and still links to the downloadable base and instruction-tuned models. [1][2]

Model access
Downloadable base and instruction-tuned v0.1 weights. Mistral’s hosted v0.1 offering retired on March 30, 2025. [1]
License
Apache License 2.0 for the published model weights. [1][2]

What the model does

On September 27, 2023, Mistral AI released its first open-weight model. It made a base model available for text generation and released a separate instruction-tuned version designed to respond to prompts in a conversational format. The release included code that developers could use as a starting point for running the weights locally. [1][2]

Why it matters

The downloadable files gave small teams a practical way to test a language model on infrastructure they controlled. Mistral’s launch tests also showed that a model of this size could compete with larger models on selected English and coding tasks, making size one factor among several when choosing a model. [2]

Who can use it

Software teams evaluating self-hosted text assistants and researchers comparing language models can use the original weights as a reference point. Teams maintaining older v0.1 integrations have separate decisions to make about hosted service migration and their own deployments. [1][2]

Key dates

  1. Model weights released [1][2]
  2. Hosted v0.1 retirement [1]

Ways to use it

Choose the base or instruction-tuned checkpoint for the task you want to test. Budget for your own hosting, application safeguards and evaluation. Mistral’s documentation still provides public weight links and directs new hosted integrations to Ministral 3 8B. [1]

Model limitations

The original model has an 8K-token context window. Mistral’s launch benchmarks reflect the 2023 tasks it tested. The surrounding application needs its own moderation for the original instruction-tuned model. [1][2]

Sources and verification

Catalog evidence checked How we verify records

  1. Primary source Mistral 7B v0.1 model card and retirement date
  2. Primary source Mistral 7B
  3. Independent source TechCrunch on the Mistral 7B open-weight release

Related profiles

Models & assistants

GPT-6.2 Astra and Sol

OpenAI

Astra targets demanding projects. Sol targets everyday tasks. Both support text and image input in the API.

High confidenceEvidence checked

Models & assistants

MAI Code 2 and Code 2 Flash

Microsoft AI

Code 2 targets complex software work. Flash targets frequent edits, explanations and routine fixes.

High confidenceEvidence checked

Models & assistants

MAI Realtime 1

Microsoft AI

Users can try a MAI Labs demo. Developers can register for a limited Foundry preview.

High confidenceEvidence checked