LLMQ: Efficient lower-precision LLM training for consumer GPUs PMLR Schultheis, Erik Alistarh, Dan-Adrian ; https://orcid.org/0000-0003-3650-940X ddc:000 We present LLMQ, an end-to-end CUDA/C++ implementation for medium-sized language-model training, e.g. 3B to 32B parameters, on affordable, commodity GPUs. These devices are characterized by low memory availability and slow communication compared to datacentre-grade GPUs. Consequently, we showcase a range of optimizations that target these bottlenecks, including activation checkpointing, offloading, and copy-engine based collectives. LLMQ is able to train or fine-tune a 7B model on a single 16GB mid-range gaming card, or a 32B model on a workstation equipped with 4 RTX 4090s. This is achieved while executing a standard 8-bit training pipeline, without additional algorithmic approximations, and maintaining FLOP utilization of around 50%. The efficiency of LLMQ rivals that of production-scale systems on much more expensive cloud-grade GPUs. ML Research Press 2026 info:eu-repo/semantics/conferenceObject doc-type:conferenceObject text http://purl.org/coar/resource_type/c_5794 https://research-explorer.ista.ac.at/record/21932 https://research-explorer.ista.ac.at/download/21932/21942 Schultheis E, Alistarh D-A. LLMQ: Efficient lower-precision LLM training for consumer GPUs. In: <i>2nd Conference on Parsimony and Learning</i>. Vol 328. ML Research Press; 2026:265-284. eng info:eu-repo/semantics/altIdentifier/e-issn/2640-3498 info:eu-repo/semantics/openAccess