jejwalsh/EVE-Instruct-GGUF

EVE-Instruct GGUF

GGUF quantizations of eve-esa/EVE-Instruct, produced with llama.cpp using an importance matrix for improved quality.

EVE-Instruct is a 24B Mistral Small 3.2 finetune specialised for Earth Observation and Earth Science, created by ESA Phi-lab, Pi School, and Mistral AI.

Available Quantizations

File	Quant	Size
eve_v05-Q4_K_M.gguf	Q4_K_M	14.3 GB
eve_v05-Q5_K_M.gguf	Q5_K_M	16.8 GB
eve_v05-Q6_K.gguf	Q6_K	19.3 GB
eve_v05-Q8_0.gguf	Q8_0	25.1 GB

Usage

llama.cpp

llama-cli -m eve_v05-Q5_K_M.gguf -ngl 99 -c 8192 -p "Your prompt"

Ollama

# Create a Modelfile
echo 'FROM ./eve_v05-Q5_K_M.gguf' > Modelfile
ollama create eve -f Modelfile
ollama run eve

LM Studio

Download any of the GGUF files above and load directly in LM Studio.