Deci AI, founded in Tel Aviv, builds deep-learning optimization technology that makes models faster and cheaper to run in production.
Deci’s platform automatically improves neural-network architectures and compiles them for efficient inference, delivering large speedups on standard hardware. Its tools cover both vision models and LLM inference โ including its DeciLM family and the Infery inference engine โ helping teams cut serving costs without sacrificing accuracy. Now part of NVIDIA, Deci’s technology continues to serve teams deploying AI under real latency and budget constraints.
Deci offers developer free access for evaluation, with commercial licenses for production inference and enterprise optimization engagements.
ML platform teams and engineers who must serve models within strict latency and cost budgets on existing hardware.
Deci helps teams whose models work in notebooks but fail their latency and cost budgets in production.
Submit models to the optimization platform, compare benchmarks on target hardware and deploy with the Infery engine.
Multiple-fold inference speedups on existing hardware, directly translating into lower serving bills or faster products.
Models are submitted for optimization, benchmarked on target hardware and deployed through the Infery engine with integration support.
Speedups vary by architecture and hardware, and teams must validate accuracy after compression.
Platform engineers report meaningful latency and cost reductions on workloads they thought were already tuned.
Inference cost is the silent killer of AI products. Deci’s optimization squeezes far more from hardware you already own. Platform teams serving high-volume models should benchmark before buying more GPUs.
Benchmark Deci before buying more compute. For most serving stacks, optimization delivers faster savings than any procurement negotiation.
Free access is available for evaluation and small workloads; production licensing and enterprise optimization are paid.
Yes โ optimized models can be deployed through Deci’s inference engine and APIs integrated into serving infrastructure.
Against serving on raw frameworks, Deci delivers measured speedups and cost reductions through architecture search and compilation.
Supabase is an open-source Firebase alternative built on Postgres, providing database, auth, storage, and real-time APIs for developers. OverviewSupabase gives developers a…
Groq runs open models at record-breaking speed on its custom LPU inference chips, delivering the fastest token throughput available for AI applications…
Mem0 is a universal memory layer for AI agents and assistants, giving applications persistent, personalized memory that improves every interaction.OverviewMem0 sits between…
