How much VRAM does embeddinggemma 300M need?

embeddinggemma 300M (0.30000001192092896B parameters) requires approximately 7.0 GB of memory with Q6_K quantization.

What is the best quantization for embeddinggemma 300M?

The recommended quantization for embeddinggemma 300M is Q6_K, which balances quality and memory efficiency.

Can it run?

Yes, NVIDIA L40 48GB can run embeddinggemma 300M with a C grade (Runs well). Expected decode speed: 653.2 tok/s.

CUsable

Runs well

Using Q6_K in Ollama

Capabilities:

Fit status

Runs well

Decode

653.2 tok/s

TTFT

350 ms

Safe context

109K

Memory

7.0 GB / 48.0 GB

Weights0.2 GB

KV Cache0.8 GB

Runtime1.2 GB

Headroom4.8 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	C	Runs well	707.6 tok/s	398 ms	218K
Chat	C	Runs well	707.6 tok/s	350 ms	54K
Coding	C	Runs well	653.2 tok/s	350 ms	109K
RAG	C	Runs well	707.6 tok/s	497 ms	218K
Reasoning	C	Runs well	707.6 tok/s	350 ms	109K

How embeddinggemma 300M (0.30000001192092896B params) fits at each quantization level on NVIDIA L40 48GB (48.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	0.1 GB	Low	D29
Q3_K_S	3	0.1 GB	Low	D29
NVFP4	4

HuggingFace

huggingface-cli download hf-ggml-org--embeddinggemma-300m-gguf

Upgrade options

MacBook Pro M4 Max 96GBBudget pick

C361.2 tok/s decode

~$2,499 MSRP

MacBook Pro M3 Max 128GBBest value

C252 tok/s decode

~$2,499 MSRP