google/embeddinggemma-300m
Primitive: /encode · Encode ·
Gemma 3
Model Page: EmbeddingGemma
Dense
Overview
Hardware: — drives latency, throughput & cost
| Size | 303M params |
|---|---|
| Tasks | /encode |
| License | gemma |
| Latency | 87 ms |
| Throughput | 27.2K tok/s |
| Cost | $0.0082 /1M tok |
Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.
Embedding
| Output types | Dense |
|---|---|
| Dimensions | dense: 768 |
| Max sequence length | 2,048 |
| Inputs | text |
Benchmarks
CQADupstackPhysicsRetrieval
Duplicate question retrieval from StackExchange Physics
Corpus: 38,314 Queries: 1,039
Quality
ndcg at 10 0.4733
map at 10 0.4090
mrr at 10 0.4670
CosQA
Code search with natural language queries
Corpus: 6,267 Queries: 500
Quality
ndcg at 10 0.3981
map at 10 0.3067
mrr at 10 0.3095
FiQA2018
Financial opinion mining and question answering
Corpus: 57,599 Queries: 648
Quality
ndcg at 10 0.2649
map at 10 0.2011
mrr at 10 0.3240
LegalBenchConsumerContractsQA
Question answering on consumer contracts
Corpus: 153 Queries: 396
Quality
ndcg at 10 0.8032
map at 10 0.7513
mrr at 10 0.7500
NFCorpus
Biomedical literature search from NutritionFacts.org
Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.2619
map at 10 0.0815
mrr at 10 0.4417
Performance L4 b1 c16
Corpus 27.2K tok/s
Corpus p50 86.8ms
Query 540 tok/s
Query p50 82.0ms
Performance RTX-4090 b1 c16
Corpus 79.6K tok/s
Corpus p50 55.7ms
Query 1.9K tok/s
Query p50 27.8ms
SCIDOCS
Citation prediction, document classification, and recommendation for scientific papers
Corpus: 25,656 Queries: 1,000
Quality
ndcg at 10 0.1092
map at 10 0.0598
mrr at 10 0.1956
SciFact
Scientific claim verification using research literature
Corpus: 5,183 Queries: 300
Quality
ndcg at 10 0.7054
map at 10 0.6565
mrr at 10 0.6670
StackOverflowQA
Programming question answering from Stack Overflow
Corpus: 19,931 Queries: 1,994
Quality
ndcg at 10 0.6961
map at 10 0.6744
mrr at 10 0.6744
Compare (0)Compare →