Clover-35B-A3B-Preview (IQ3_XXS GGUF)
This repository contains an IQ3_XXS quantisation of Diffuspace/Clover-35B-A3B-Preview for use with llama.cpp, ollama, and other GGUF-compatible runtimes.
Clover is a 35B-parameter mixture-of-experts (MoE) language model with roughly 3B active parameters per token, derived from the Qwen3.6-35B-A3B architecture. It features a 128K context window and block-diffusion generation[reference:0].
Quantisation Details
| Parameter | Value |
|---|---|
| Quantisation type | IQ3_XXS (≈3.06 bits per weight) |
| Importance matrix | Yes (imatrix_unsloth.gguf) |
| Output tensor type | q8_0 |
| Token embedding type | q8_0 |
| Source model | Clover-35B-A3B-Preview-Q8_0.gguf |
| Quantiser | llama.cpp build b4838ffb (Linux x86-64, CPU) |
Exact quantisation command
llama-clover-b4838ffb-linux-x64-cpu/bin/llama-quantize \
--allow-requantize \
--output-tensor-type q8_0 \
--token-embedding-type q8_0 \
--imatrix imatrix_unsloth.gguf \
Clover-35B-A3B-Preview-Q8_0.gguf \
Clover-35B-A3B-Preview-IQ3_XXS.gguf \
IQ3_XXS