Running Massive Models Locally on Low-End Hardware
Felix_455-788 · reddit · 2026-07-12
The author runs 100B to 200B+ large models on a low-spec laptop, relying heavily on MoE models, quantization, mmap, and offloading some parameters to NVMe/CPU to maintain usability.
They primarily use LM Studio, running models like Deepseek-V4-Flash (UD-IQ3XXS) and Nemotron-3-Super-120B-A12B (UD-Q3KM) at speeds around 1.0–2.5 tok/s. Because they use it in a "batch processing" manner—sending a task and moving on to other things—they find it sufficient for reverse engineering, bug fixing, and code auditing.
Finally, they note they mainly use opencode and Pi with context lengths between 16K–90K. Due to payment restrictions in their country, relying on cloud models long-term isn't an option, so they will continue using local LLMs in extreme scenarios.
More from Infra
- oMLX 0.5.2 adds Mac menu-bar stats, low-bit decode kernels, and faster downloads — awnihannun · 2026-07-22
- Strangeworks launches Aura to turn enterprise ops into production optimization systems — whurley · 2026-07-22
- Graph workload 854.graph500 enters SPEC CPU 2026 as a new CPU benchmark — Prof_DavidBader · 2026-07-22
- Hybrid and local inference are emerging as a response to AI energy and token costs — dmitry140 · 2026-07-22
- NVIDIA details Vera CPU with 2x performance claims and a 22,000-core rack — ryanshrout · 2026-07-22
- NVIDIA says Vera Rubin NVL72 delivers 10x more tokens per megawatt than Blackwell — nvidia · 2026-07-22