community pool / model answer

VibeThinker-3B

chat · WeiboAI/VibeThinker-3B

46.3 tok/s · model median · 30 runs in the model record.

Reference rig	Quantization	Median result	Basis	TTFT (ms)	Peak VRAM	Max context	Engines
RTX 3060 12GB	4-bit	45.9 tok/s	measured	648	- GB	131,072	llama.cpp
RTX 3060 12GB ×2	16-bit	45.7 tok/s	measured	2,727	7.9 GB	131,072	llama.cpp
RTX 5000 16GB	8-bit	76.6 tok/s	reported	-	6.1 GB	65,536	llama.cpp
RX 570 4GB ×2	4-bit	51.1 tok/s	reported	52	- GB	2,048	llama.cpp
RX 570 4GB ×2	5-bit	46.6 tok/s	reported	32	- GB	2,048	llama.cpp

Context tested

The table above is the tested context for each exact rig and quantization cell. Empty fields remain empty rather than being filled with an estimate.