CanItRun Logocanitrun.

Gemma 4 E2B

Gemma 4 E2B needs roughly 3.2 GB VRAM at Q4_K_M quantization (10.5 GB at FP16). 117 GPUs we track can run it fully in VRAM at 8k context.

117 GPUs run this natively · 1 with CPU offload

Google4.65B params128k contextApache 2.0Commercial use ok

Gemma 4 E2B is a 4.65B parameter dense model developed by Google. Released 2 April 2026 under Apache 2.0, the smaller of Google's two on-device Gemma 4 models. The E stands for effective parameters: 2.3B run per token, but the checkpoint stores 5.1B once its Per-Layer Embedding (PLE) tables are counted (the text-only GGUF holds 4.65B), and those tables must sit in memory. 35 layers in a repeating pattern of four 512-token sliding-window layers and one full-attention layer, with the last 20 layers sharing earlier layers' KV cache. Text, image, and audio input at 128K context.

To run Gemma 4 E2B locally: Size the card for the 4.65B of loaded weights, not the 2B in the name: Q4_K_M weights are about 2.8 GB and the total is 3.24 GB at 8k context, and Q8_0 is 5.60 GB in total, so it runs on a 6 GB card at Q8_0 only barely. Published GGUF files agree: Q4_K_M is 3.11 GB from unsloth and 3.46 GB from bartowski, and Q8_0 is about 5.0 GB. Only 15 layers keep a KV cache, so the full 131,072-token window adds just 0.81 GB. Decode speed follows the 2.3B effective parameters, which is why it is fast on entry-level GPUs and CPUs. Vision and audio load from a separate mmproj file; Ollama's gemma4:e2b download is 4.6 GB.

Google reports MMLU-Pro 60.0, GPQA Diamond 43.4, and LiveCodeBench v6 44.0, high for a model that runs 2.3B parameters per token. These are vendor-measured figures from the model card.

VRAM at each quantization

Gemma 4 E2B natively supports a longer context window, but the table below is capped at 8k for comparability; its hybrid attention stack keeps KV cache growth well below linear.

QuantWeightsKV cacheTotal
FP3218.6 GB0.06 GB20.9 GB
BF169.3 GB0.06 GB10.5 GB
FP169.3 GB0.06 GB10.5 GB
Q8_0rec4.9 GB0.06 GB5.6 GB
Q6_K3.8 GB0.06 GB4.3 GB
Q5_K_M3.3 GB0.06 GB3.8 GB
Q4_K_M2.8 GB0.06 GB3.2 GB
Q3_K_M2.2 GB0.06 GB2.6 GB
Q2_K1.8 GB0.06 GB2.0 GB
NVFP4cuda2.3 GB0.06 GB2.7 GB

KV cache figures assume 8k context at FP16. NVFP4 quantization requires a CUDA-capable GPU. Enable TurboQuant in the calculator to see reduced KV cache estimates.

Benchmarks

MMLU-Pro
60.0
GPQA Diamond
43.4
LiveCodeBench
44.0

GPUs that run Gemma 4 E2B natively (117)

Show 112 more
Plus 1 GPU that runs it with CPU offload (slower)

No GPU? Rent an RTX 3080 10GB from $0.17/hr ↗

Prices as of Oct 17, 2026. Rates change; check RunPod for current pricing. Affiliate link: CanItRun may earn a commission. Doesn't affect the fit calculation above.

Notes

The E stands for effective parameters: 2.3B run per token, but about 4.65B of text weights load (Per-Layer Embedding tables are resident yet only looked up), and the audio and vision encoders add roughly 0.45B more in the separate mmproj file. Ollama's Q4_K_M is a 4.6 GB download for that reason. Sliding-window attention (512 tokens) plus KV sharing across the last 20 layers keeps the long-context cache tiny.

Hugging Face ↗Ollama ↗Released 2026-04-02

Continue reading

vram-guides8 min
Best LLMs for 6 GB VRAM (2026)

Frequently asked questions

What are the VRAM requirements for Gemma 4 E2B?
Gemma 4 E2B requires approximately 3.2 GB of VRAM at Q4_K_M quantization, 5.6 GB at Q8, and 10.5 GB at FP16. These numbers assume 8k context window; its hybrid attention stack caches far fewer than all layers, so VRAM grows much slower than linearly with context.
How many parameters does Gemma 4 E2B have?
Gemma 4 E2B has 4.65 billion parameters.
How capable is Gemma 4 E2B?
With an MMLU-Pro score of 60, Gemma 4 E2B delivers solid general-purpose performance suitable for most everyday tasks and professional use.
Can Gemma 4 E2B run on a 16 GB GPU?
Yes. Gemma 4 E2B needs 3.2 GB at Q4_K_M, which fits in a 16 GB GPU like the RTX 4080 or RTX 5070 Ti.
What is the smallest quantization for Gemma 4 E2B that fits in 24 GB of VRAM?
At BF16, Gemma 4 E2B needs 10.5 GB, the highest-quality quantization that fits in 24 GB of VRAM.
What GPU do I need to run Gemma 4 E2B locally?
A 16 GB GPU is enough. At Q4_K_M, Gemma 4 E2B needs 3.2 GB VRAM. Good options: RTX 4080 (16 GB), RTX 5070 Ti (16 GB).
Can I run Gemma 4 E2B in the cloud?
Yes. The cheapest RunPod setup we found that holds Gemma 4 E2B entirely in GPU memory at 8k context is 1× RTX 3080 10GB at Q8_0, at roughly 171.2 tokens per second and from $0.17 per hour for the whole pod. Prices as of Oct 17, 2026; RunPod rates change, so check their pricing page for the current figure.