Google · google/gemma-4-12b
| Provider | |
|---|---|
| Family | gemma |
| Type | vlm |
| Status | active |
| Released | 2026-05-23 |
| Updated | 2026-07-15 |
| Parameters | 12B |
| Open weights | yes |
Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.
| Device | Memory | Bandwidth | Best quality | Weights | tok/s | Fastest |
|---|---|---|---|---|---|---|
| Cerebras WSE-3 | 44.0 GB | 21000000.0 GB/s | bf16 | 23.92 GB | ~614562.4 | ~2458249.4 q4 |
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | bf16 | 23.92 GB | ~468.2 | ~1873.0 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | bf16 | 23.92 GB | ~234.1 | ~936.5 q4 |
| NVIDIA B200 | 180.0 GB | 7700.0 GB/s | bf16 | 23.92 GB | ~225.3 | ~901.4 q4 |
| Google TPU7x (Ironwood) | 192.0 GB | 7380.0 GB/s | bf16 | 23.92 GB | ~216.0 | ~863.9 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | bf16 | 23.92 GB | ~175.6 | ~702.4 q4 |
| AMD Instinct MI300X | 192.0 GB | 5300.0 GB/s | bf16 | 23.92 GB | ~155.1 | ~620.4 q4 |
| NVIDIA H200 SXM | 141.0 GB | 4800.0 GB/s | bf16 | 23.92 GB | ~140.5 | ~561.9 q4 |
| NVIDIA H100 NVL | 94.0 GB | 3900.0 GB/s | bf16 | 23.92 GB | ~114.1 | ~456.5 q4 |
| NVIDIA H100 SXM | 80.0 GB | 3350.0 GB/s | bf16 | 23.92 GB | ~98.0 | ~392.1 q4 |
| AMD Instinct MI250X | 128.0 GB | 3200.0 GB/s | bf16 | 23.92 GB | ~93.6 | ~374.6 q4 |
| Google TPU v5p | 95.0 GB | 2765.0 GB/s | bf16 | 23.92 GB | ~80.9 | ~323.7 q4 |
| NVIDIA A100 80GB SXM | 80.0 GB | 2039.0 GB/s | bf16 | 23.92 GB | ~59.7 | ~238.7 q4 |
| NVIDIA H100 PCIe | 80.0 GB | 2000.0 GB/s | bf16 | 23.92 GB | ~58.5 | ~234.1 q4 |
| NVIDIA GeForce RTX 5090 | 32.0 GB | 1792.0 GB/s | bf16 | 23.92 GB | ~52.4 | ~209.8 q4 |
| Google TPU v6e (Trillium) | 32.0 GB | 1638.0 GB/s | bf16 | 23.92 GB | ~47.9 | ~191.7 q4 |
| AMD Instinct MI210 | 64.0 GB | 1600.0 GB/s | bf16 | 23.92 GB | ~46.8 | ~187.3 q4 |
| NVIDIA A100 40GB SXM | 40.0 GB | 1555.0 GB/s | bf16 | 23.92 GB | ~45.5 | ~182.0 q4 |
| Google TPU v4 | 32.0 GB | 1200.0 GB/s | bf16 | 23.92 GB | ~35.1 | ~140.5 q4 |
| NVIDIA GeForce RTX 4090 | 24.0 GB | 1008.0 GB/s | fp8 | 11.96 GB | ~59.0 | ~118.0 q4 |
| AMD Radeon RX 7900 XTX | 24.0 GB | 960.0 GB/s | fp8 | 11.96 GB | ~56.2 | ~112.4 q4 |
| NVIDIA GeForce RTX 5080 | 16.0 GB | 960.0 GB/s | fp8 | 11.96 GB | ~56.2 | ~112.4 q4 |
| NVIDIA GeForce RTX 3090 | 24.0 GB | 936.0 GB/s | fp8 | 11.96 GB | ~54.8 | ~109.6 q4 |
| NVIDIA L40S | 48.0 GB | 864.0 GB/s | bf16 | 23.92 GB | ~25.3 | ~101.1 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | bf16 | 23.92 GB | ~24.0 | ~95.9 q4 |
| AMD Radeon RX 7900 XT | 20.0 GB | 800.0 GB/s | fp8 | 11.96 GB | ~46.8 | ~93.6 q4 |
| Apple M2 Ultra | 192.0 GB | 800.0 GB/s | bf16 | 23.92 GB | ~23.4 | ~93.6 q4 |
| Google TPU v5e | 16.0 GB | 800.0 GB/s | fp8 | 11.96 GB | ~46.8 | ~93.6 q4 |
| NVIDIA GeForce RTX 4080 SUPER | 16.0 GB | 736.0 GB/s | fp8 | 11.96 GB | ~43.1 | ~86.2 q4 |
| NVIDIA GeForce RTX 4070 Ti SUPER | 16.0 GB | 672.0 GB/s | fp8 | 11.96 GB | ~39.3 | ~78.7 q4 |
| AMD Radeon RX 9070 XT | 16.0 GB | 640.0 GB/s | fp8 | 11.96 GB | ~37.5 | ~74.9 q4 |
| Apple M4 Max | 128.0 GB | 546.0 GB/s | bf16 | 23.92 GB | ~16.0 | ~63.9 q4 |
| Apple M1 Max | 64.0 GB | 400.0 GB/s | bf16 | 23.92 GB | ~11.7 | ~46.8 q4 |
| Apple M2 Max | 96.0 GB | 400.0 GB/s | bf16 | 23.92 GB | ~11.7 | ~46.8 q4 |
| Apple M3 Max | 128.0 GB | 400.0 GB/s | bf16 | 23.92 GB | ~11.7 | ~46.8 q4 |
| NVIDIA GeForce RTX 3060 12GB | 12.0 GB | 360.0 GB/s | q6 | 8.97 GB | ~28.1 | ~42.1 q4 |
| NVIDIA L4 | 24.0 GB | 300.0 GB/s | fp8 | 11.96 GB | ~17.6 | ~35.1 q4 |
| NVIDIA GeForce RTX 4060 Ti 16GB | 16.0 GB | 288.0 GB/s | fp8 | 11.96 GB | ~16.9 | ~33.7 q4 |
| Apple M4 Pro | 64.0 GB | 273.0 GB/s | bf16 | 23.92 GB | ~8.0 | ~32.0 q4 |
| Apple M4 | 32.0 GB | 120.0 GB/s | bf16 | 23.92 GB | ~3.5 | ~14.0 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Hugging Face | google/gemma-4-12B |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|