Can a AMD Instinct MI300X (192GB) run Gemma 3 12B?

Yes — Gemma 3 12B runs at FP16 (23.5 GB) at about 112–169 tok/s on a AMD Instinct MI300X (192GB).

How much memory does Gemma 3 12B need?

Gemma 3 12B needs about 6.2 GB at 4K context at its smallest quantization. A AMD Instinct MI300X (192GB) has 192.0 GB.

Yes — Gemma 3 12B runs on a AMD Instinct MI300X (192GB) at FP16

Yes — it runs great.

Runs great

Best fit: Gemma 3 12B at FP16 · 23.5 GB weights · 112–169 tok/s on a AMD Instinct MI300X (192GB).

Arena Text Overall · style-controlled

1,342FP16 untested

gemma-3-12b-it · published configuration reference · FP16 not tested

Arena Text Overall (style-controlled), exact published entry gemma-3-12b-it, 2026-10-08: reference rating 1,341.8. Published confidence bounds 1,332.3–1,351.3. 3,829 votes. Published gemma-3-12b-it configuration reference. Reasoning mode is not disclosed by the Arena dataset. Tested precision, runtime and sampling settings are not disclosed in this dataset. Selected FP16 has not been measured by this Arena entry. No quantization quality adjustment is applied.

Snapshot published 2026-10-08 · snapshot imported 2026-10-11 UTC · validated last-good data · manual refresh

Arena leaderboard dataset · Arena / lmarena-ai, Arena Leaderboard Dataset (CC BY 4.0). Selected Text Overall style-controlled rows; field names adapted, ratings unchanged. · dataset CC BY 4.0 · revision e16f90232986

Total memory at 4K context: 24.9 GB; Weights + KV cache + runtime overhead
Estimated generation: 112–169 tok/s
Estimated first token: ~0.6s

Every quantization on this machine

On small screens, each quant lists every metric below its name. On wider screens, scroll sideways for total memory, speed and fit. Keyboard: focus the table, then use the arrow keys. Quant names stay visible.

Gemma 3 12B quantizations on AMD Instinct MI300X (192GB), estimated at 4K context. Weights and total memory are shown separately.
Quant	Weights	Total memory	Speed	Fit
Q2_K	4.8 GB	6.2 GB	244–366 tok/s	Runs great
Q3_K_M	6.0 GB	7.4 GB	226–340 tok/s	Runs great
Q4_K_M	7.3 GB	8.7 GB	211–316 tok/s	Runs great
Q5_K_M	8.5 GB	9.9 GB	198–298 tok/s	Runs great
Q6_K	9.7 GB	11.1 GB	187–280 tok/s	Runs great
Q8_0	12.5 GB	13.9 GB	165–247 tok/s	Runs great
FP16	23.5 GB	24.9 GB	112–169 tok/s	Runs great

Open the full calculator →Everything a AMD Instinct MI300X (192GB) runs Gemma 3 12B on other machines What to buy to run it →

Ran Gemma 3 12B on a AMD Instinct MI300X (192GB)? Add your real numbers →