Quants
Collection
148 items • Updated • 3
How to use from
llama.cppInstall from WinGet (Windows)
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16# Run inference directly in the terminal:
llama cli -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16Use pre-built binary
# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16# Run inference directly in the terminal:
./llama-cli -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16Build from source code
git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16# Run inference directly in the terminal:
./build/bin/llama-cli -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16Use Docker
docker model run hf.co/neopolita/mistral-nemo-minitron-8b-base-gguf:F16Quick Links
GGUF quants for nvidia/Mistral-NeMo-Minitron-8B-Base using llama.cpp
Terms of Use: Please check the original model
Quants
q2_k: Uses Q4_K for the attention.vw and feed_forward.w2 tensors, Q2_K for the other tensors.q3_k_s: Uses Q3_K for all tensorsq3_k_m: Uses Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_Kq3_k_l: Uses Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_Kq4_0: Original quant method, 4-bit.q4_1: Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models.q4_k_s: Uses Q4_K for all tensorsq4_k_m: Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q4_Kq5_0: Higher accuracy, higher resource usage and slower inference.q5_1: Even higher accuracy, resource usage and slower inference.q5_k_s: Uses Q5_K for all tensorsq5_k_m: Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q5_Kq6_k: Uses Q8_K for all tensorsq8_0: Almost indistinguishable from float16. High resource use and slow. Not recommended for most users.
- Downloads last month
- 8
Hardware compatibility
Log In to add your hardware
16-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Install (macOS, Linux)
# Start a local OpenAI-compatible server with a web UI: llama serve -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16# Run inference directly in the terminal: llama cli -hf neopolita/mistral-nemo-minitron-8b-base-gguf:F16