FMplex: New System for Efficiently Serving Multiple Foundation Models Through Shared Backbones
Researchers have introduced FMplex, a serving system that allows multiple AI downstream tasks to share a single foundation model backbone rather than each running a separate model instance. Current deployment practice replicates large model weights for every task, wasting accelerator memory and forfeiting batching efficiencies. FMplex addresses a growing infrastructure bottleneck as foundation models proliferate across industry applications.
FMplex, presented in a preprint submitted to arXiv on June 8, 2026, proposes treating foundation model (FM) backbones as a virtualization substrate analogous to how operating systems virtualize hardware resources. Each downstream task receives a 'virtual foundation model' (vFM) — a logically private instance backed by a shared physical model — preserving task-specific extensions and independent lifecycles while eliminating redundant backbone copies. The system also introduces a batch-aware fair-queueing scheduler that combines weighted task-level sharing with both inter- and intra-task batching across co-located tasks. Evaluated across 7 FM backbones (16 variants) and 92 downstream tasks spanning language, vision, time-series, and multimodal domains, FMplex reduced inference latency by up to 80% compared to spatial partitioning and 33.3% compared to best-effort co-location. At cluster scale, the system was able to host up to six times more tasks than baseline approaches. The work targets a practical pain point in production AI infrastructure, where the cost of deploying many fine-tuned or extended model variants on limited GPU memory is substantial.
What's missing
The paper is a preprint and has not yet undergone peer review. Key open questions include how FMplex handles backbone version updates or model heterogeneity across tasks, whether the fairness guarantees of the scheduler hold under adversarial or highly skewed workloads, and how overhead scales when tasks require substantially different backbone modifications (e.g., architectural extensions vs. adapter layers). Real-world production deployment results are not reported.
What different sources said
- arXiv cs.AICenter
FMplex: Model Virtualization for Serving Extensible Foundation Models
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.