sloppy

Slopsome brings LLM & GPU stats straight into Discord — no setup, no API key.

Ask whether a model will run on a given GPU and get a real answer: VRAM fit (in-VRAM / with offload / needs more), plus prefill (compute-bound) and decode (bandwidth-bound) tokens/sec — for any quant: GGUF k-quants & i-quants, AWQ, GPTQ, EXL2/EXL3.

Commands

• /fit — will a model run on a GPU? VRAM + tok/s, any quant & context
• /model — specs, benchmarks, per-quant VRAM, pricing for one model
• /gpu — VRAM, bandwidth, TFLOPS and reference tok/s for a GPU
• /leaderboard — top models by composite score
• /search — find a model by name / maker / family
• /help — list everything

Covers local open-weight models and hosted API models. Data comes from slopsome.com — a free, no-login search engine for LLM & GPU stats. Invite the bot and type /help.