Open-source Colibri runs GLM-5.2 744B MoE on a 25GB RAM laptop
socialwithaayan · x · 2026-07-21
- An open-source pure-C runtime called colibri claims it can run GLM-5.2, a 744B MoE model, on a machine with about 25GB of RAM.
- The key idea is that only about 40B parameters are active per token, while experts are streamed from disk and a learning cache keeps the hot set resident.
- The post shows measured speeds ranging from 0.05 tok/s on a 25GB dev box to 1.8 tok/s on a 128GB CPU desktop and 5.8–6.8 tok/s on a 6× RTX 5090 setup.
- The runtime is described as single-file, zero-dependency C, with token-exact validation against a full-precision oracle and a demo that reports 9.9GB resident memory after startup.
Related event: Open-Source Colibri Runs 744B GLM-5.2 on Consumer Laptops(2 posts)→
More from Infra
- RTK Terminal Compression Cuts Tokens but Leaves Your AI Coding Bill Unchanged — Bartaseth · 2026-09-11
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- PyTorch Day Korea 2026 launches first offline conf, CFP closes Sept 13 — PyTorch · 2026-09-11
- Local LLM server dilemma: 4x CMP-170HX (price up 53% in 20 days) vs Mac Studio M5 Ultra — rumboll · 2026-09-11
- llama.cpp lands Flash Attention tuning for RDNA4, big prefill gains on AMD — pmttyji · 2026-09-11