litert-community/embeddinggemma-2-740m-litert-lm

Main Model Card: google/embeddinggemma-2

EmbeddingGemma 2 740M is the full, natively multimodal variant of the EmbeddingGemma 2 family, packaged for low-latency, on-device inference using LiteRT. Integrating the 270M parameter text backbone with modular vision (170M) and audio (300M) encoders, this 740M parameter model maps text, code, images, video, and audio into a single, shared 768-dimensional vector space. It enables robust cross-modal semantics and supports interleaved multimodal inputs within a single 8,192-token context window. Built for ultimate flexibility, it allows developers to run full multimodal search, speech retrieval, and complex cross-media semantic similarity pipelines directly on consumer hardware.

Try EmbeddingGemma 2 with LiteRT-LM

Android	iOS	macOS	Web

Ready to integrate this into your product? Get started here.

Try EmbeddingGemma 2 with MediaPipe

MediaPipe uses EmbeddingGemma 2 and LiteRT-LM to enable high-level, cross-platform tasks which you can experience firsthand with the following interactive web demos:

• Semantic image search powered by MediaPipe Universal Embedder and Semantic Retriever task.
• Decision making powered by MediaPipe Decision Maker.

Model Specifications

	EmbeddingGemma 2 Text 270M	EmbeddingGemma 2 Text Vision 440M	EmbeddingGemma 2 740M
Supported Modalities	Text	Text, Images	Text, Images, Video, Audio
Parameters	Total: 270M Transformer: 130M Embeddings: 140M	Total: 440M Transformer: 130M Embeddings: 140M Vision Encoder: 170M	Total: 740M Transformer: 130M Embeddings: 140M Vision Encoder: 170M Audio Encoder: 300M
Quantization Scheme	Transformer:     int4 per-channel (QAT) Embeddings:     int4 per-channel (QAT)	Transformer:     int4 per-channel (QAT) Embeddings:     int4 per-channel (QAT) Vision Encoder:     int8 per-channel (QAT)	Transformer:     int4 per-channel (QAT) Embeddings:     int4 per-channel (QAT) Vision Encoder:     int8 per-channel (QAT) Audio Encoder:     mixed int2/int4/int8     per-channel (QAT)
Download Size (CPU/GPU file)	165 MB	388 MB	485 MB
On-Demand Modality Loading	-	Yes	Yes
Supported Input Sizes	Text tokens: 128, 256, 512,     1024, 2048, 8192	Text tokens: 128, 256, 512,     1024, 2048, 8192 Image soft tokens: 70, 140	Text tokens: 128, 256, 512,     1024, 2048, 8192 Image soft tokens: 70, 140 Audio soft tokens: 12
Supported Output Sizes	128 (512 bytes) 256 (1024 bytes) 512 (2048 bytes) 768 (3072 bytes)	128 (512 bytes) 256 (1024 bytes) 512 (2048 bytes) 768 (3072 bytes)	128 (512 bytes) 256 (1024 bytes) 512 (2048 bytes) 768 (3072 bytes)
Link	litert-community/embeddinggemma-2-text-270m-litert-lm	litert-community/embeddinggemma-2-text-vision-440m-litert-lm	litert-community/embeddinggemma-2-740m-litert-lm

Additional Notes:

• LiteRT-LM automatically scales and patchifies images of any size to fit EmbeddingGemma 2's vision encoder model.
• LiteRT-LM supports EmbeddingGemma 2's streamed audio allowing a variety of different audio lengths.

EmbeddingGemma 2 Performance on LiteRT-LM

The text performance was measured by loading and running the 128 text signature. For vision benchmarking, the vision encoder used the 70 signature. The audio file for the Text + Audio run was 10 seconds long while the file for Text + Audio + Vision was 5 seconds. The latency reported is the average of 5 iterations.

Memory was measured with each platform's native metrics and is not directly comparable across operating systems. CPU memory was measured using, rusage::ru_maxrss on Android, Linux and IOT, task_vm_info::phys_footprint on iOS and MacBook, and process_memory_counters::PrivateUsage on Windows. With the exception of task_vm_info::phys_footprint, accelerator (GPU/NPU/TPU) memory is not included. Memory measurements were taken from the second load which reads from loading caches. Memory usage on the first load may vary.

Android

Device	Backend	Text latency	Text + Vision latency	Text + Audio latency	Text + Vision + Audio latency
Google Pixel 11 Pro	TPU	8.3 ms	49 ms	193.5 ms	149 ms
S26 Ultra	CPU	27.1 ms	175 ms	305 ms	322 ms
S26 Ultra	GPU	25.9 ms	119 ms	350 ms	324 ms

Device	Backend	Text CPU memory	Text + Vision CPU memory	Text + Audio CPU memory	Text + Vision + Audio CPU memory
Google Pixel 11 Pro	TPU	112 MB	125 MB	121 MB	127.5 MB
S26 Ultra	CPU	334 MB	694 MB	465 MB	811 MB
S26 Ultra	GPU	333 MB	427 MB	407 MB	493 MB

iOS

Device	Backend	Text latency	Text + Vision latency	Text + Audio latency	Text + Vision + Audio latency
iPhone 18 Pro	CPU	41.8 ms	191 ms	445 ms	400 ms
iPhone 18 Pro	GPU	11.6 ms	69.8 ms	165 ms	150 ms

Device	Backend	Text CPU/GPU memory	Text + Vision CPU/GPU memory	Text + Audio CPU/GPU memory	Text + Vision + Audio CPU/GPU memory
iPhone 18 Pro	CPU	84 MB	196 MB	98 MB	196 MB
iPhone 18 Pro	GPU	85 MB	196 MB	96 MB	226 MB

Linux

Device	Backend	Text latency	Text + Vision latency	Text + Audio latency	Text + Vision + Audio latency
Arm 2.3 & 2.8GHz	CPU	105 ms	825 ms	947 ms	1251 ms
NVIDIA GeForce RTX 4090	GPU	7.6 ms	23.9 ms	169 ms	112 ms

Device	Backend	Text CPU memory	Text + Vision CPU memory	Text + Audio CPU memory	Text + Vision + Audio CPU memory
Arm 2.3 & 2.8GHz	CPU	310 MB	647 MB	431 MB	764 MB
NVIDIA GeForce RTX 4090	GPU	528 MB	678 MB	654 MB	811 MB

macOS

Device	Backend	Text latency	Text + Vision latency	Text + Audio latency	Text + Vision + Audio latency
MacBook Pro M5	CPU	31.3 ms	151 ms	314 ms	432 ms
MacBook Pro M5	GPU	9.5 ms	37.3 ms	195 ms	131 ms

Device	Backend	Text CPU/GPU memory	Text + Vision CPU/GPU memory	Text + Audio CPU/GPU memory	Text + Vision + Audio CPU/GPU memory
MacBook Pro M5	CPU	165 MB	310 MB	201 MB	336 MB
MacBook Pro M5	GPU	233 MB	403 MB	320 MB	478 MB

Windows

Device	Backend	Text latency	Text + Vision latency	Text + Audio latency	Text + Vision + Audio latency
Dell XPS 16 (Intel Core Ultra Series 3)	CPU	71.7 ms	338 ms	813 ms	708 ms
Dell XPS 16 (Intel Core Ultra Series 3)	GPU	19.2 ms	62.5 ms	284 ms	205 ms
Dell XPS 16 (Intel Core Ultra Series 3)	Running Intel OpenVINO NPU	13.3 ms	49.8 ms	124.3 ms	108.6 ms

Device	Backend	Text CPU memory	Text + Vision CPU memory	Text + Audio CPU memory	Text + Vision + Audio CPU memory
Dell XPS 16 (Intel Core Ultra Series 3)	CPU	211 MB	342 MB	242 MB	369 MB
Dell XPS 16 (Intel Core Ultra Series 3)	GPU	822 MB	1769 MB	1047 MB	1962 MB
Dell XPS 16 (Intel Core Ultra Series 3)	Running Intel OpenVINO NPU	201 MB	394 MB	288 MB	503 MB

Web

Device	Backend	Text latency	Text + Vision latency	Text + Audio latency	Text + Vision + Audio latency
MacBook Pro M5	GPU	21.8 ms	107 ms	228 ms	226 ms

IoT

Device	Backend	Text latency	Text + Vision latency	Text + Audio latency	Text + Vision + Audio latency
Raspberry Pi 5 16GB	CPU	161 ms	1761 ms	946 ms	2304 ms
Jetson Orin Nano	CPU	207 ms	1741 ms	1659 ms	2507 ms
Jetson Orin Nano	GPU	88 ms	485 ms	1195 ms	1069 ms
Arduino VENTUNO Q	NPU	13.6 ms	135 ms	- *	- *

Device	Backend	Text CPU memory	Text + Vision CPU memory	Text + Audio CPU memory	Text + Vision + Audio CPU memory
Raspberry Pi 5 16GB	CPU	282 MB	619 MB	408 MB	736 MB
Jetson Orin Nano	CPU	302 MB	627 MB	417 MB	741 MB
Jetson Orin Nano	GPU	523 MB	1103 MB	755 MB	1228 MB
Arduino VENTUNO Q	NPU	206 MB	460 MB	- *	- *

* Audio modality is not supported on the IQ-8275 NPU so this metric is omited.