Colibri: Running a 744B Parameter Model on 25GB of RAM
Discover Colibri: a tiny C engine that streams experts from disk to run massive 744B parameter models like GLM-5.2 on machines with only 25GB of RAM.
Tag moe
Running Frontier MoE Models on Consumer Hardware via vLLM-Moet
Deploy frontier MoE models like GLM-5.2 (753B) on consumer Blackwell GPUs using 2-bit experts and FP4 recovery with the vLLM-Moet patch.
Scaling Beyond Limits: Introducing the LongCat-2.0 MoE Model
Explore LongCat-2.0, a massive 1.6 trillion parameter MoE model featuring 1M context window and superior performance on SWE-bench Pro.