Miner Guide

Hardware requirements

LLM miner

  • openhermes-2.5-mistral-7b-gptq: This is a 4-bit quantized version of a small Mistral 7b model. Mining rewards for this model is discontinued.

  • openhermes-2-pro-mistral-7b: This is a full-precision Mistral 7b model. Requires NVIDIA GPU with 24GB of VRAM (examples: RTX 3090, RTX 4090) 0.2x reward rate.

  • (Recommended) dolphin-2.9-llama3-8b: This is a full-precision Llama3 8b model. Requires NVIDIA GPU with 24GB of VRAM (examples: RTX 3090, RTX 4090) 0.24x reward rate.

  • (Recommended) openhermes-mixtral-8x7b-gptq: This is a 4-bit quantized version of Mixtral 8x7b model. Requires NVIDIA GPU with 40GB of VRAM (examples: RTX A6000, A100-40GB) 1x reward rate.

  • openhermes-2-yi-34b-gptq: This is a 8-bit quantized version of Yi-34b model. Requires NVIDIA GPU with 40GB of VRAM (examples: RTX A6000, A100-40GB) 1x reward rate.

Stable Diffusion miner

  • NVIDIA cards with at least 12GB of VRAM

Why do we require such high-end GPUs? Read this article: https://heuristai.medium.com/heurists-guide-to-ai-beginner-s-series-part-2-db77458a62dd

Official Installation guide

See https://github.com/heurist-network/miner-release

Community-made guide

Our worldwide mining community members have created many tutorials.

PreviousHeurist LLM GatewayNextModel Hosting Guide

Last updated