The open source control plane for AI inference
Joined June 2026
- Tweets11
- Following1
- Followers19
- Likes8
Introducing Inference Day. 🚀
A free, live virtual event on the future of AI infrastructure. Nov 5, 8 to 11 AM PT.
@kelseyhightower, @vfarcic, and builders from NVIDIA, Red Hat, Tetrate, Ericsson, Qualcomm and Dash0.
Save your spot 👉 modelplane.ai/inference-day
Modelplane retweeted
Big release today with integration with @NVIDIAAI . We’ve been working with Nvidia on this and excited to release the work.
A powerful story - NVIDIA Dynamo runs inference on the cluster, AICR ensures it’s well matched to the hardware, and Modelplane turns into a global inference service.
The stack is a per-cluster choice in @Modelplane v0.4, and it's immutable, so adoption is incremental.
Stand up a @NVIDIA Dynamo cluster next to the ones you already run. Same ModelDeployment on both.
Blog link: modelplane.ai/blog/modelplan…
The stack is a per-cluster choice in @Modelplane v0.4, and it's immutable, so adoption is incremental.
Stand up a @NVIDIA Dynamo cluster next to the ones you already run. Same ModelDeployment on both.
Blog link: modelplane.ai/blog/modelplan…
Compute isn't the bottleneck. The missing control loop is. @CoreWeave runs @kubernetesio all the way down. No VMs, just containers and a reconcile loop.
That's not a coincidence, it's the only pattern that survives at AI scale. @bassamtabbara explains on Hello, Agent! 👇🎧
The largest inference operations share one pattern: a declarative reconcile loop.
@bassamtabbara, Founder and CEO at @upbound_io, points to CoreWeave, running Kubernetes to no VMs, just containers.
Click here for the "Hello, Agent!" episode. bit.ly/4gUOAv6
#RealTimeData #DataStreaming #Redpanda
Modelplane v0.3 is out.
Run Claude Code on your own GPUs: we now serve the Anthropic Messages API end to end. Plus @Vultr as a fifth cluster provider, per-cluster credentials for multi-account fleets, and a local test that runs with no cloud and no GPU.
→ modelplane.ai/blog/modelplan…
"Day 0 support" usually means someone dropped everything. Nemotron 3.5 Lightning was serving on Modelplane the same afternoon @nvidia shipped it and nothing in Modelplane changed. We don't wrap engines or pin versions.
modelplane.ai/blog/nemotron-…
#vLLM
We measured our Claude Code usage at API prices: 13x our seat cost on average, 52x for our heaviest engineer.
Nothing broke. That's Claude Code working as intended, once you see the meter.
What's your subsidy multiple? → modelplane.ai/blog/ai-coding…
Cluster-level inference tooling is fine. It just doesn't help you orchestrate the GPUs you own.
Nobody at scale runs one cluster. The fix: treat the ones you own as one pool.
That's the fleet-level API we designed.
How we built it →modelplane.ai/blog/how-we-de…
Modelplane v0.2 is out.
Two new providers (@Nebius, @Azure AKS), weighted routing for safe model rollouts, and cluster taints to reserve or drain capacity.
The models, engines, clouds, and accelerators you already run, operated as one system. →modelplane.ai/blog/modelplan…
Modelplane retweeted
looking forward to seeing this evolve on top of @crossplane_io and more
Today we’re open sourcing Modelplane, the open source control plane for AI inference.
Open-weight models changed who runs AI. As inference expands across clouds, accelerators, and regions, operators need a fleet layer above the cluster.
That’s @ModelplaneAI. Join us.
modelplane.ai/blog/open-cont…
Modelplane is here: an open source control plane for AI inference. Any model, any engine, any infra, from one GPU to multi-node.
It's v0.1 and built in the open. Spin it up on a kind cluster, then tell us where it breaks. ⭐ → modelplane.ai