Mistral-7B-Instruct-v0.2
Mistral AIMistral-7B-Instruct-v0.2 is an instruction-tuned 7B chat model with a 32K context window, fine-tuned from the Mistral-7B-v0.2 base model and prompted with [INST] tags. Mistral describes it as a quick demonstration that the base model can be fine-tuned easily, and the card notes it has no moderation mechanisms.
Superseded by Mistral-7B-Instruct-v0.3. The card carries no benchmark table and links only the base-model paper, so the capability figures come from the paper and the La Plateforme launch post.
Openness
3 high confidence- weights
- open(Apache-2.0, on HF)
- data
- closed
- code
- closed(no training/post-training pipeline)
- license
- Apache-2.0(OSI permissive)
Apache-2.0 open weights, with the instruction-tuning data and recipe undisclosed, so this is open weights rather than open source.
- https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2 recorded 2026-09-29
Hugging Face model page. The card says "instruct fine-tuned version of the Mistral-7B-v0.2" and the repo tags include "license:apache-2.0"; the embedded state has "gated":false (weights ungated).
- https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2/raw/main/README.md recorded 2026-09-29
Raw model card. It states "The Mistral-7B-Instruct-v0.2 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-7B-v0.2." and points to the paper and release blog for details. Searched the body for dataset, training data, training code and fine-tuning recipe: none published.
Adoption
4 high confidenceAdoption is read from Hugging Face downloads on the single declared SKU, Mistral-7B-Instruct-v0.2, which also carries a large derivative ecosystem of finetunes, adapters and merges. The checkpoint remains heavily used as a permissive 7B instruct base years after its release.
- https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2 recorded 2026-08-13
1,206,680 downloads in the trailing 30 days; model tree reads 1,100 finetunes / 1,200 adapters / 351 merges
Capability
2 medium confidenceMistral reports 7.6 on MT-Bench for v0.2, and the Mistral 7B paper gives 6.84 for the original v0.1 instruct model, which it says outperformed all 7B models on that benchmark. Later Mistral 7B releases have replaced this version.
- https://arxiv.org/html/2310.06825 recorded 2026-09-29
Mistral 7B paper (HTML full text). The instruction-tuning section says "exhibits superior performance compared to all 7B models on MT-Bench" for the original Mistral 7B Instruct (v0.1; 6.84 in the table), and the base-model comparison table is captioned "Comparison of Mistral 7B with Llama"; neither table covers the v0.2 instruct tune.
- https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2 recorded 2026-09-29
Model card page. It describes the model as an "instruct fine-tuned version of the Mistral-7B-v0.2" with a "32k context window (vs 8k context in v0.1)" and names Mistral-7B-Instruct-v0.3 as the newer version.
- https://mistral.ai/news/la-plateforme recorded 2026-09-29
La Plateforme launch post: "Mistral-tiny only works in English. It obtains 7.6 on MT-Bench." and "a new minor release of Mistral 7B Instruct".
Verified 2026-09-27