BEFORE YOU DOWNLOAD

Found a model file? Check the fit.

Use the size of your chosen download, rather than an approximate catalogue weight size. Keep the model architecture and context matched to your setup.

No upload needed. Enter the combined size of all weight shards for a supported GGUF or native MXFP4 download. This replaces the catalogue weight allowance only. Match the exact model architecture; compressed archives, adapters and formats expanded by the runtime are not suitable.

Match your download

Exact model architectureCombined weight file sizeFile-size unitContext tokensComputer typeGPU memory (GiB)System RAM (GiB)

FILE-BASED PLANNING ESTIMATE

Within your memory budget

7.4 GiB

GPU memory target, including cache and headroom.

Entered weights: 4.66 GiB
FP16 attention cache: 1.00 GiB
System RAM target: 16 GiB

A file-based estimate does not verify format support or guarantee a successful load. One text conversation, full GPU offload and a compatible Ollama or LM Studio backend are assumed.

GB means one billion bytes; GiB means 1,073,741,824 bytes. Include separately loaded weight components where relevant. Vision processing, multiple users and CPU offload are not calculated. See this model's architecture and limitations.