Spark-X2.5-1.7B-Base — imatrix Q4_K_M

Q4_K_M GGUF quantization made from the author's BF16 GGUF with a combined importance matrix derived from all three CSV files in the calibration dataset.

Artifacts

Spark-X2.5-1.7B-Base-imatrix-Q4_K_M.gguf — 1,107,458,112 bytes
combined.imatrix.gguf — 1,918,784 bytes
manifest.json — provenance, source hashes, and output hashes
smoke-test.txt — validation record

The combined matrix contains statistics for 168 tensors computed across 27 512-token chunks. Use a Spark-X2.5-compatible build from XHToken/llama.cpp; generic builds may not yet recognize this architecture.

Downloads last month: 72

GGUF

Model size

473k params

Architecture

Hardware compatibility

4-bit

View +4 variants

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for fuzzy-mittenz/Spark-X2.5-1.7B-Base-imatrix-Q4_K_M

Base model

XHToken/Spark-X2.5-1.7B-Base

Quantized

(8)

this model