How Tensormesh Platform works with LMCache, which workloads benefit, compatibility, pricing, security and support.
LMCache is the open-source KV cache engine created by our team, and it powers Tensormesh’s core caching technology. Tensormesh is the platform built on top of it: a complement to your inference stack with advanced caching techniques, monitoring, reporting, optimization, and support. If you want to run and tune LMCache yourself, the open-source project is free. If you want the performance gains and our help to optimize your infrastructure, Tensormesh Platform is the product.
Caching pays off when requests share or repeat context. That includes multi-turn chat, teamwork over the same documents (including but not limited to RAG), agents that reuse long system prompts or tool definitions, and long-context workloads. Workloads where every request is short and unique will see smaller gains, because there is less to reuse. If you are in doubt, you can contact us for a free evaluation of your workload.
Results depend on how much context your workload reuses. In controlled environments we have measured up to 10x performance improvement. The platform shows your actual performance, helps you gauge your savings, and helps you optimize them.
Please see our compatibility page.
Yes. Tensormesh Platform supports peer-to-peer (P2P) KV cache sharing, so nodes in a cluster can reuse cache computed elsewhere instead of recomputing it. This boosts cache hit rates across your whole deployment, not just on a single GPU.
Tensormesh Platform is designed to fit into existing AI infrastructure, and we work with partners across storage and enterprise platforms. See our partners page or contact us to talk about your environment.
A cache miss happens when the system looks for data in the cache and it isn’t there, so it has to fetch or recompute it the slow way.
Pricing is tailored to your deployment and workload. Contact our sales team for a quote.
Tensormesh Platform runs on your premises and none of your data is shared with Tensormesh. Some anonymous usage data is shared with the LMCache project but can be deactivated. See the usage stats documentation.
If you are a Tensormesh Platform customer, our support team is available via Slack or email. Our subscription agreement offers 24h support during business hours with a 4h response time.
Didn’t find your answer?
