DCP
Open Navigation

Docs

ENAR

Mistral 7B Instruct v0.2

Fast and cost-efficient 7B instruction model deployment guide.

1. What it is

Mistral 7B Instruct v0.2 (`mistralai/Mistral-7B-Instruct-v0.2`) is a 7B instruction-tuned model from Mistral AI.

2. What it does

It provides strong general chat, summarization, and assistant behavior with good latency/cost efficiency.

3. How it compares

  • Versus TinyLlama/Phi-2: better response quality for a moderate VRAM increase.
  • Versus ALLaM/Falcon/JAIS: usually cheaper/faster, but Arabic-specialized models can outperform on Arabic-heavy domains.

4. Best for on DCP

  • General-purpose bilingual copilots
  • Budget-sensitive production chat APIs
  • Summarization and support automation

5. Hardware requirements on DCP

  • DCP floor: `min_vram_gb: 16`
  • Recommended providers: RTX 3090/4090
  • Template: `vllm-serve`

6. How to run on DCP

  1. Use `job_type: "vllm_serve"`.
  2. Set `params.model: "mistralai/Mistral-7B-Instruct-v0.2"`.
  3. Set `gpu_requirements.min_vram_gb` to `16`.

7. Licensing and commercial-use notes

Mistral 7B Instruct v0.2 is published under Apache-2.0 on Hugging Face and is generally usable in commercial systems with standard notice obligations.

Sources:

  • https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2
  • /home/node/dc1-platform/docker-templates/vllm-serve.json
  • /home/node/dc1-platform/backend/src/db.js