Run zai-org/GLM-5.3-Flash as an EXL3 4bpw checkpoint on NVIDIA DGX Spark (GB10) with vLLM and cuda-exl3 — measured two-node (TP=2) and three-node (TP=3 + expert parallel) recipes, plus the kernel fixes, NCCL mesh patches, loader work and benchmarks underneath them, and what we retracted.
Run zai-org/GLM-5.3-Flash (NVFP4) on 3x NVIDIA DGX Spark with vLLM, TP=3 + EP, DFlash2 speculative decoding, CUDA graphs. Full recipe with rationale, benchmarks, patches, and what we tried.
GLM-5.3 Flash EXL3 for 2x DGX Sparks
DeepSeek-v4-Flash 0731 recipe for 2x DGX Sparks
The open source coding agent.
Define, Prompt and Test MCP enabled Agents and Workflows
Documentation site for fast-agent