Mistral 7B v0.1
Mistral 7B v0.1 provides a pretrained text model and an instruction-tuned companion. Both use Apache 2.0. The base model is a foundation for further training rather than a chat assistant by default.
The original architecture
The base card identifies grouped-query attention, sliding-window attention and a byte-fallback BPE tokenizer. The sliding window reduces attention work compared with attending to the full sequence at every layer. This distinguishes v0.1 from the later v0.2 architecture.Mistral 7B v0.1 primary source
Base and Instruct serve different roles
Mistral-7B-Instruct-v0.1 is fine-tuned on public conversation datasets and uses INST delimiters. Mistral-7B-v0.1 is the pretrained base checkpoint. Neither card supplies built-in moderation as a complete application safety system. The two checkpoints share this release article.Mistral 7B v0.1 official model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| Model identifiers | Mistral-7B-v0.1Mistral-7B-Instruct-v0.1 |
| Context window | 8,000 tokens The model catalog advertises 8k context for this version. This is not a measured recall limit. Mistral 7B v0.1 model and lifecycle record |
| Weight license | Apache 2.0Mistral 7B v0.1 official model card |
| Access and lifecycle | Hosted API retired on 30 March 2025. Downloadable weights remain listed.Mistral 7B v0.1 model and lifecycle recordMistral 7B v0.1 official model card |
| Checkpoint variants | Mistral-7B-v0.1 and Mistral-7B-Instruct-v0.1Mistral 7B v0.1 official model cardMistral 7B v0.1 primary source |