gemma-4-12B-it

12BChatby Google Start a discussion

0

Quantized builds

42

Benchmark records

Quantized builds

Community and official quantized releases; "linked records" counts only benchmarks tied to that exact release.

No quantized builds cataloged yet.

Benchmark data

42 configurations

HardwareFrameworkQuantDecodeVRAM usedSamplesEvidence
NVIDIA RTX 5060 SUPERvLLMQ8_029.3 tok/s12 GB1L0 Self-reported
NVIDIA RTX 5060 SUPERvLLMQ5_K_M42.8 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5060 SUPERvLLMQ4_K_M52.2 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5060 SUPERllama.cppQ8_026.9 tok/s12 GB1L0 Self-reported
NVIDIA RTX 5060 SUPERllama.cppQ5_K_M39.3 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5060 SUPERllama.cppQ4_K_M47.9 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5070 SUPERvLLMQ8_058.6 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 5070 SUPERvLLMQ5_K_M85.6 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5070 SUPERvLLMQ4_K_M104.3 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5070 SUPERllama.cppQ8_053.8 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 5070 SUPERllama.cppQ5_K_M78.5 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5070 SUPERllama.cppQ4_K_M95.7 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 3090vLLMQ8_061.2 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 3090vLLMQ5_K_M89.4 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 3090vLLMQ4_K_M109 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppQ8_056.2 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppQ5_K_M82 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppQ4_K_M100 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5070 TivLLMQ8_058.6 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 5070 TivLLMQ5_K_M85.6 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5070 TivLLMQ4_K_M104.3 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppQ8_053.8 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppQ5_K_M78.5 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppQ4_K_M95.7 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5070vLLMQ8_043.9 tok/s12 GB1L0 Self-reported
NVIDIA RTX 5070vLLMQ5_K_M64.2 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5070vLLMQ4_K_M78.2 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5070llama.cppQ8_040.3 tok/s12 GB1L0 Self-reported
NVIDIA RTX 5070llama.cppQ5_K_M58.9 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5070llama.cppQ4_K_M71.8 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5090vLLMQ8_0117.2 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 5090vLLMQ5_K_M171.1 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5090vLLMQ4_K_M208.7 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppQ8_0107.5 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppQ5_K_M157.1 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppQ4_K_M191.5 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 4090vLLMQ8_065.9 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 4090vLLMQ5_K_M96.3 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 4090vLLMQ4_K_M117.4 tok/s7.8 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppQ8_060.5 tok/s13.5 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppQ5_K_M88.3 tok/s9.4 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppQ4_K_M107.7 tok/s7.8 GB1L0 Self-reported

42 of these records are not linked to a specific quantized build; the full set is on the search page.

Community discussion

No discussions yet — start the first one.

Start a discussion