community pool / model answer

Qwen3-Coder-30B-A3B-Instruct

code · Qwen/Qwen3-Coder-30B-A3B-Instruct

80.4 tok/s · model median · 21 runs in the model record.

Reference rig	Quantization	Median result	Basis	TTFT (ms)	Peak VRAM	Max context	Engines
Arc Pro B70 32GB ×2	4-bit	93.3 tok/s	reported	226	- GB	2,048	llama.cpp
M1 Max 64GB	4-bit	72.3 tok/s	reported	24	- GB	2,048	ollama
M3 Ultra 512GB	4-bit	100 tok/s	reported	79	- GB	8,192	llama.cpp
Radeon AI Pro R9700 32GB ×3	8-bit	79 tok/s	reported	204	- GB	686	llama.cpp
RX 7900 XTX 24GB	4-bit	43.8 tok/s	reported	2,116	24.2 GB	2,048	llama.cpp
Ryzen AI Max 395 128GB	5-bit	80.4 tok/s	measured	-	23.9 GB	100,000	llama.cpp
Ryzen AI Max 395 64GB	4-bit	31.9 tok/s	measured	699	30.3 GB	131,072	llama.cpp
Tesla V100 32GB ×2	6-bit	100.5 tok/s	reported	-	58 GB	65,536	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.