Gemma 4 12B

57 records across 14 hardware.

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX PRO 6000 Blackwell Workstation EditionGemma 4 12B Q4_K_Mllama.cpp287.33––––2026-10-12sourced–source
RTX PRO 6000 Blackwell Workstation EditionGemma 4 12B Q4_K_Mllama.cpp285.98––––2026-10-12sourced–source
RTX 5090Gemma 4 12B Q4_K_Mllama.cpp279.09––––2026-10-12sourced–source
RTX 5090Gemma 4 12B Q4_K_Mllama.cpp278.37––––2026-10-12sourced–source
Arc B580Gemma 4 12B QATllama.cpp (SYCL)100––––2026-09community–source
GB10Gemma 4 12B Q4_K_Mllama.cpp68.42––––2026-10-12sourced–source
GB10Gemma 4 12B Q4_K_Mllama.cpp67.31––––2026-10-12sourced–source
RX 6800Gemma 4 12B Q4_K_Mllama.cpp (Vulkan)52.1799––tg1282026-10-01community–source
RX 6800Gemma 4 12B Q4_K_Mllama.cpp (ROCm)48.9736––tg1282026-10-01community–source
Arc B580Gemma 4 12B QATllama.cpp (SYCL)40––––2026-09community–source
Radeon RX 9060 XTGemma 4 12B Q4_K_Mllama.cpp (Vulkan)39.01509––tg1282026-10-02community–source
M5 MaxGemma 4 12B 8bitoMLX38.6943.1––10242026-10-07community–source
M5 MaxGemma 4 12B 8bitoMLX38.61181.0––10242026-10-09community–source
M5 MaxGemma 4 12B 8bitoMLX38.51589.0––40962026-10-09community–source
M5 MaxGemma 4 12B 8bitoMLX38.31505.0––40962026-10-11community–source
M5 MaxGemma 4 12B 8bitoMLX38.2907.3––10242026-10-11community–source
M5 MaxGemma 4 12B 8bitoMLX38.01433.0––40962026-10-07community–source
Radeon RX 9060 XTGemma 4 12B Q4_K_Mllama.cpp (ROCm)37.61467––tg1282026-10-02community–source
M5 MaxGemma 4 12B 8bitoMLX37.61224.0––40962026-10-17community–source
M5 MaxGemma 4 12B 8bitoMLX37.41610.0––163842026-10-17community–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM32.7––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM32.62––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM31.41––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM31.38––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM28.8––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM28.59––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM28.09––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM28.09––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM26.82––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM26.64––––2026-10-11sourced–source
M4 MaxGemma 4 12B 8bitoMLX25.1412.4––163842026-10-11community–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM24.79––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM24.57––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM23.91––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM23.6––––2026-10-11sourced–source
M5 ProGemma 4 12B nvfp4oMLX23.0956.6––40962026-10-16community–source
M5 ProGemma 4 12B nvfp4oMLX23.0907.9––163842026-10-16community–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM20.88––––2026-10-11sourced–source
Radeon AI PRO R9700Gemma 4 12B FP8vLLM20.65––––2026-10-11sourced–source
M6Gemma 4 12B 12BQATLM Studio19––––2026-10community–source
M4 ProGemma 4 12B 8bitoMLX17.3274.5––10242026-10-07community–source
M4 ProGemma 4 12B 8bitoMLX17.2271.6––40962026-10-07community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)13.93–––40962026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)13.77–––40962026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)13.33–––163842026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)12.58–––327682026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)12.15–––163842026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)10.66–––327682026-10communitycontradictionsource
M3 ProGemma 4 12B 8bitoMLX10.2213.8––10242026-10-14community–source
M3 ProGemma 4 12B 8bitoMLX10.2212.7––10242026-10-14community–source
M3 ProGemma 4 12B 8bitoMLX10.1216.5––10242026-10-14community–source
M3 ProGemma 4 12B 8bitoMLX10.1210.9––10242026-10-14community–source
M3 ProGemma 4 12B 8bitoMLX10.1215.2––40962026-10-14community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)–261.5–––2026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)–263.2–––2026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)–186.7––163842026-10community–source
Adreno X2-90Gemma 4 12B Q4_0llama.cpp (OpenCL)–209.3––163842026-10community–source

← all models