Lumabri
AI
distributed llm inference
SHOW HN2026
How HN took it
4 positive, 4 neutral or asking, 1 mixed, 3 critical, out of 12 comments about the project. Thread score +0.20.
Execution: 0 for, 3 againstIdea: 2 for, 0 againstPrivacy: 1 for, 0 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Interesting idea for distributed local inference
From 3 comments
This looks really interesting, and if I understand what this does properly: it was high time someone built this! Without diving into an experiment myself, it would be amazing if…
SCHiM on HNI’ve been thinking about this same idea recently, so I’m glad it exists now! The biggest benefit I see is to enable RAM constrained GPUs to perform inference of large parameter…
gebdev on HNGreat work. Love the idea. I have been thinking along the same way, but for training. For inference, the waiting time might be turn off for users
bicepjai on HN
Useful for running models on a LAN with spare hardware
From 2 comments
I am sorry I did not understand all of it. But, would this allow running large MoE LLMs on a local network with experts spread out over multiple cheaper GPUs (or even CPUs)? This…
brainless on HNThat’s one of the strongest use-cases. On a local network (office, lab, home cluster) the RTT is a few milliseconds instead of 20-50 ms, so the expert-offloading becomes much…
vforno on HN
Objections
Floating-point results may not be deterministic across hardware, weakening verification
From 3 comments
> Expert execution is deterministic (pure matmul). Isn't that only true in theory but wrong in practice due to floating points?
stymaar on HNThe harder case isn't same-hardware determinism; it's that lumabri wants CPU and GPU peers in the same swarm. Accumulation order differs across hardware, not just across runs.
nikunjbjj on HNThere are fixed-point models too which can be deterministic. But for floating point you are very instruction-set dependent, never mind floating point operations not being…
snovv_crash on HN
Latency over the internet may be high
From 1 comment
I’ve been thinking about this same idea recently, so I’m glad it exists now! The biggest benefit I see is to enable RAM constrained GPUs to perform inference of large parameter…
gebdev on HN
Requests
Add performance stats and experiment logs
From 1 comment
This looks really interesting, and if I understand what this does properly: it was high time someone built this! Without diving into an experiment myself, it would be amazing if…
SCHiM on HN
Link explicit Hugging Face model repos instead of vague model names
From 1 comment
so looking at the start docs, you should integrating the same llamacpp setup where they just bake in huggingface support. second, don't use vague names of models; point to the…
cyanydeez on HN
Questions
How is temperature handled in verification
From 1 comment
Cool idea. How do you handle temperature in the verification?
snovv_crash on HN
Alternatives named
meshllm
From 1 comment
On the surface it seems similar to https://meshllm.cloud/, but now I notice that I don't understand either.
krautsauer on HN