A local 8B model made a GGUF quant cheat sheet — all 16 numbers check out
Impressive-Kick293 · reddit · 2026-11-22
A Reddit user generated a one-page GGUF quantization cheat sheet with SenseNova-U1.5-8B-MoT running locally (MacBook M5 Pro, MLX build, 33 seconds per 1760x2368 card) — and all 16 size-table numbers match llama.cpp's quantize README exactly.
The workflow: feed an infographic skill the README numbers, let it pick layout and style, write a long art-directed prompt, render, then have a vision model review each take — four takes per card, keeping the cleanest. Two more cards from the same session (heat pump basics, pour-over standards) also came out factually solid.
Only the image rendering is local: orchestration ran on Hermes via DeepSeek V4.1 Flash, with prompt writing and reviews routed to Gemini 3.1 Flash Lite and Qwen3-VL through OpenRouter. The skill is open-sourced as sn-infographic.
More from coding & agent
- Hugging Face's Lewis Tunstall Maps the Road from LLMs to Agent Swarms — _lewtun · 2026-11-22
- Grok Build CLI vs AFK Pilot vs T3 Code: feature comparison across three grok runners — PawelHuryn · 2026-11-22
- GPT-6.2 is hard to steer via system prompt — OpenAI says use it directly in Codex — pvncher · 2026-11-22
- Comet cut Opik dataset ingestion time from 39 minutes to 80 seconds by fixing 8 bottlenecks — dl_weekly · 2026-11-22
- Ghost Mode: open-source Rust Code Mode cuts ~60% of agent tool-call token costs — legend-of-alan · 2026-11-22
- Hangar: open-source macOS/Linux/Windows VMs to sandbox AI coding agents — ozgursoy · 2026-11-22