modelfit

Will this Hugging Face model run on my hardware, how fast, and with what command? Architecture-aware: MoE offload, hybrid/linear attention, MLA, custom runtimes. Every number shows its basis and confidence.

best coding model on one RTX 3090 + 64GB RAM what fits a 128GB DGX Spark with long context vision model for an M4 Mac with 32GB can Edge0/Edge0-120B-q3k run on a 3090 with 64gb ram at 128k Qwen/Qwen3.8-27B on an RTX 4090 at 64k context

Estimates come from repo files, tensor headers and configs (no weights downloaded). Speeds are roofline estimates with error bands until measured. MCP endpoint: /mcp · API: POST /api/ask, GET /api/estimate?repo=&hw=&ctx=