Blog
Notes from the GPU floor
Case studies, pilot results and engineering notes on running self-hosted LLMs efficiently on your own GPUs.
- Pilot
NeuroBlock pilot: 3.2× structured output throughput on the same GPUs
First results from our pilot with NeuroBlock. Structured output went from 193 to 614 tokens per second on the same NVIDIA GPUs and model, with sub-second time to first token.
- Case study
Half the monthly GPU cost on a production LLM deployment
Feniria cut monthly GPU cost by 50% on a production LLM inference workload, with p99 time to first token down 48% and no hardware added.
Work with us
Find out if self-hosting pays off for you.
Tell us your model, request volume, traffic pattern and current cost. We'll come back with an initial estimate, and an honest answer if an API is still the better choice.