Smeltcore
Which AI models actually run on your GPU — benchmarks & 800 install recipes
About
Smeltcore answers one question: will this AI model actually run on your GPU?
Read full description
Smeltcore answers one question: will this AI model actually run on your GPU? We track 84 open-weights models — LLMs, image, video, TTS, multimodal, 3D — across 27 consumer GPUs, from an RTX 3060 to Apple Silicon Macs. Every model×GPU pair gets an honest verdict (runs / tight fit / out of memory) backed by real community-measured numbers: generation speed, peak VRAM, quantization used. The hard part of local AI isn't picking a model — it's the evening you lose to CUDA versions, wrong quants and OOM crashes. So each verdict links to one of 800 step-by-step install recipes (llama.cpp, Ollama, ComfyUI, vLLM, MLX) written for the exact card you own: VRAM budget, which quant to pick, the flags that matter and the traps to avoid. There's a GPU Advisor ("what can my card run?"), side-by-side GPU comparison, and a community submission pipeline — every benchmark is source-linked and reviewed before publishing. The whole dataset is free, no signup, CC BY-SA.
- Category
- Engineering & Development, LLMs, Platforms
- Available on
- web
- Launched
- July 8, 2026
Reviews
No reviews yet. Be the first to share a useful, honest impression.
Sign in to review this app.
Opt in to release updates in your Rocket inbox. Unfollow here or in Notifications to stop future updates.




