
★ Pick of the week · WK 47 / 2026
Pro-grade 4K image generation and editing on your own GPU, free
Open weights, 16 reference images, and community builds that fit a 16 GB card.
01 / Language & code models
Let your coding agent hammer a fake AWS instead of your real oneAWS's own open-source emulator fakes 200+ services locally, so agent mistakes cost nothing.
02 / Language & code models
Same GPU, same models, up to twice the tokens per secondComfyUI 0.45 and llama.cpp 1.0 speed up local LLMs, though ComfyUI's numbers are its own.
03 / Language & code models
Near-top coding and agent quality at $3/$15, without Carbon's slownessGemini 4 Neon matched Carbon on Artificial Analysis and debuted second on Agent Arena.
04 / Language & code models
A very cheap multimodal agent model, and the weights are outHy4 Flash costs $0.07 per million input tokens. The 441 GB weights are for big rigs only.
05 / Language & code models
A free-tier Gemini one point behind Sonnet on Artificial AnalysisGemini 4 Hydrogen is free in AI Studio and hallucinated just 8% on one test.02
The State of AI
What to try first in each area, and what's hot.03
For your setup
New AI, sorted by what you can use tonight.| Model | Fit | |
|---|---|---|
04
AI models & tools
Everything worth knowing about, with what it does, what it needs and where to try it.Filters
| Name | Latest | |
|---|---|---|