Can RTX 5090 32GB run Gemma 4 12B?
Measured results for Gemma 4 12B (11.96B dense) on RTX 5090 32GB, plus what fits and the commands to run it.
Short answer
Yes. Gemma 4 12B Q8_0 fits fully on RTX 5090 32GB at 16k context, using 13.3 of 31.0 GB usable, and generates about 103.7 tokens/s in our own run. As the context fills it slows down: 99.3 tok/s with 16k tokens already in context (96% of the empty-context speed), measured. For Q8_0, our bandwidth-only estimate was ~78 tok/s against a measured median of ~104 (25% too pessimistic).
Measured generation speed by quantization
| Quant | Generation tok/s | Range | Prompt tok/s | Runs | Origin |
|---|---|---|---|---|---|
| Q4 class (median) | 152.4 | 152.4–152.4 | 8524 | 1 | includes our runs |
| Q6 class (median) | 122.3 | 122.3–122.3 | 7160 | 1 | public runs |
| Q8 class (median) | 103.7 | 103.7–103.7 | 8806 | 1 | includes our runs |
| Q4_K_M (our run, depth 0) | 152.4 | 3 repeats | 8524 | 3 | own measurement |
| Q8_0 (our run, depth 0) | 103.7 | 3 repeats | 8806 | 3 | own measurement |
Generation = tg128 (128 generated tokens, short context). Public rows are medians of community llama.cpp results (sources); our own runs use llama-bench b11658 on a rented GPU.
Speed vs context depth measured
Our own llama-bench runs on GeForce RTX 5090 (flash attention on, 3 repeats): generation and prompt speed with the given number of tokens already in the context.
| Quant | Context depth | Generation tok/s | vs depth 0 | Prompt tok/s |
|---|---|---|---|---|
| Q4_K_M | 0 | 152.4 | 100% | 8524 |
| Q4_K_M | 4k | 145.2 | 95% | 7894 |
| Q4_K_M | 16k | 143.2 | 94% | 6818 |
| Q8_0 | 0 | 103.7 | 100% | 8806 |
| Q8_0 | 4k | 100.1 | 97% | 8095 |
| Q8_0 | 16k | 99.3 | 96% | 6922 |
Alternatives
Less memory that still fits Gemma 4 12B
Smaller memory usually means a cheaper card or Mac (we do not list prices).
- Mac mini / iMac M4 16GB: Q4_K_M, 11 GB usable, ~11 tok/s scaled from measured
- RTX 5070 12GB: Q6_K, 11 GB usable, ~51 tok/s scaled from measured
- RTX 4070 12GB: Q6_K, 11 GB usable, ~43 tok/s scaled from measured
Other models on RTX 5090 32GB, similar size
- Qwen3.5 9B: Q8_0, fits fully, ~156 tok/s (measured)
- Ornith 1.5 9B (hardware page): Q8_0, fits fully, ~145 tok/s (scaled from measured)
- Qwen3-VL 8B: Q8_0, fits fully, ~167 tok/s (measured)
- Gemma 4 E4B (hardware page): Q8_0, fits fully, ~177 tok/s (scaled from measured)
Memory and context limits
Q8_0 weights are 12.7 GB; the 16k f16 KV cache adds 0.60 GB. Largest context that fits in 31.0 GB usable:
| KV cache type | Max context |
|---|---|
| f16 | 256k |
| q8_0 | 256k |
| q4_0 | 256k |
Native limit 256k. Full KV cache table for Gemma 4 12B.
Copy-paste commands
The advisor's pick for this hardware: Q8_0, 16k context.
llama-server -hf unsloth/gemma-4-12b-it-GGUF:Q8_0 -c 16384 -fa on --jinja -ngl 99
OLLAMA_CONTEXT_LENGTH=16384 ollama serve ollama run hf.co/unsloth/gemma-4-12b-it-GGUF:Q8_0
FAQ
How fast is Gemma 4 12B on RTX 5090 32GB?
About 103.7 tokens/s for generation (Q8_0 on llama.cpp CUDA, 128-token generation), 8806 tokens/s for prompt processing. Speed drops as the context fills; see the depth table.
How much memory does Gemma 4 12B need on RTX 5090 32GB?
The advisor's pick, Q8_0, is 12.7 GB of weights plus 0.60 GB of KV cache at 16k context (13.3 GB), against 31.0 GB usable.
Is the Gemma 4 12B speed on RTX 5090 32GB measured or estimated?
Measured: our own llama-bench run on a rented GeForce RTX 5090, including runs with 4k to 16k tokens already in context.
What can I use instead of RTX 5090 32GB for Gemma 4 12B?
With less memory, Mac mini / iMac M4 16GB (~11 tok/s), RTX 5070 12GB (~51 tok/s), RTX 4070 12GB (~43 tok/s) also fit it fully.
What is the maximum context for Gemma 4 12B on RTX 5090 32GB?
With an f16 KV cache about 256k tokens, 256k with q8_0 and 256k with q4_0, bounded by the model's own limit.
Related
All models on RTX 5090 32GB · Gemma 4 12B on all hardware · VRAM tier · all measured pages · leaderboard
Compare: RTX 4090 24GB vs RTX 5090 32GB · RTX 3090 24GB vs RTX 5090 32GB · 2x RTX 3090 (48GB total) vs RTX 5090 32GB