Qwen3 Embedding 4B vs Cohere Embed v3
Detailed comparison between Qwen3 Embedding 4B and Cohere Embed v3. See which embedding best meets your accuracy and performance needs.
Model Comparison
Qwen3 Embedding 4B takes the lead.
Both Qwen3 Embedding 4B and Cohere Embed v3 are powerful embedding models designed to improve retrieval quality in RAG applications. However, their performance characteristics differ in important ways.
Why Qwen3 Embedding 4B:
- Qwen3 Embedding 4B delivers better accuracy (nDCG@10: 0.802 vs 0.686)
- Cohere Embed v3 is 57172ms faster on average
Overview
Key metrics
ELO Rating
Overall ranking quality
Qwen3 Embedding 4B
Cohere Embed v3
Win Rate
Head-to-head performance
Qwen3 Embedding 4B
Cohere Embed v3
Accuracy (nDCG@10)
Ranking quality metric
Qwen3 Embedding 4B
Cohere Embed v3
Average Latency
Response time
Qwen3 Embedding 4B
Cohere Embed v3
Visual Performance Analysis
Performance
ELO Rating Comparison
Win/Loss/Tie Breakdown
Accuracy Across Datasets (nDCG@10)
Latency Distribution (ms)
Breakdown
How the models stack up
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Overall Performance | |||
| ELO Rating | 1496 | 1488 | Overall ranking quality based on pairwise comparisons |
| Win Rate | 41.1% | 41.0% | Percentage of comparisons won against other models |
| Pricing & Availability | |||
| Price per 1M tokens | $0.020 | $0.100 | Cost per million tokens processed |
| Release Date | 2025-06-06 | 2024-02-07 | Model release date |
| Accuracy Metrics | |||
| Avg nDCG@10 | 0.802 | 0.686 | Normalized discounted cumulative gain at position 10 |
| Performance Metrics | |||
| Avg Latency | 80021ms | 22849ms | Average response time across all datasets |
Dataset Performance
By field
Comprehensive comparison of accuracy metrics (nDCG, Recall) and latency percentiles for each benchmark dataset.
PG
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| Latency Metrics | |||
| Mean | 107349ms | 35998ms | Average response time |
| P50 | 105202ms | 35278ms | 50th percentile (median) |
| P90 | 123451ms | 41398ms | 90th percentile |
business reports
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| Latency Metrics | |||
| Mean | 26100ms | 4704ms | Average response time |
| P50 | 25578ms | 4610ms | 50th percentile (median) |
| P90 | 30015ms | 5410ms | 90th percentile |
DBPedia
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| nDCG@5 | 0.603 | 0.634 | Ranking quality at top 5 results |
| nDCG@10 | 0.595 | 0.619 | Ranking quality at top 10 results |
| Recall@5 | 0.234 | 0.219 | % of relevant docs in top 5 |
| Recall@10 | 0.375 | 0.353 | % of relevant docs in top 10 |
| Latency Metrics | |||
| Mean | 106747ms | 30356ms | Average response time |
| P50 | 104612ms | 29749ms | 50th percentile (median) |
| P90 | 122759ms | 34909ms | 90th percentile |
FiQa
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| nDCG@5 | 0.724 | 0.641 | Ranking quality at top 5 results |
| nDCG@10 | 0.759 | 0.650 | Ranking quality at top 10 results |
| Recall@5 | 0.715 | 0.639 | % of relevant docs in top 5 |
| Recall@10 | 0.835 | 0.678 | % of relevant docs in top 10 |
| Latency Metrics | |||
| Mean | 120137ms | 33241ms | Average response time |
| P50 | 117734ms | 32576ms | 50th percentile (median) |
| P90 | 138158ms | 38227ms | 90th percentile |
SciFact
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| nDCG@5 | 0.695 | 0.729 | Ranking quality at top 5 results |
| nDCG@10 | 0.733 | 0.769 | Ranking quality at top 10 results |
| Recall@5 | 0.787 | 0.788 | % of relevant docs in top 5 |
| Recall@10 | 0.893 | 0.900 | % of relevant docs in top 10 |
| Latency Metrics | |||
| Mean | 143712ms | 41205ms | Average response time |
| P50 | 140838ms | 40381ms | 50th percentile (median) |
| P90 | 165269ms | 47386ms | 90th percentile |
MSMARCO
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| nDCG@5 | 1.000 | 1.000 | Ranking quality at top 5 results |
| nDCG@10 | 1.000 | 0.996 | Ranking quality at top 10 results |
| Recall@5 | 0.123 | 0.123 | % of relevant docs in top 5 |
| Recall@10 | 0.224 | 0.218 | % of relevant docs in top 10 |
| Latency Metrics | |||
| Mean | 112945ms | 30494ms | Average response time |
| P50 | 110686ms | 29884ms | 50th percentile (median) |
| P90 | 129887ms | 35068ms | 90th percentile |
NorQuAD
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| Latency Metrics | |||
| Mean | 13597ms | 3703ms | Average response time |
| P50 | 13325ms | 3629ms | 50th percentile (median) |
| P90 | 15637ms | 4258ms | 90th percentile |
ARCD
| Metric | Qwen3 Embedding 4B | Cohere Embed v3 | Description |
|---|---|---|---|
| Accuracy Metrics | |||
| nDCG@5 | 0.915 | 0.349 | Ranking quality at top 5 results |
| nDCG@10 | 0.922 | 0.398 | Ranking quality at top 10 results |
| Recall@5 | 0.940 | 0.380 | % of relevant docs in top 5 |
| Recall@10 | 0.960 | 0.520 | % of relevant docs in top 10 |
| Latency Metrics | |||
| Mean | 9581ms | 3093ms | Average response time |
| P50 | 9389ms | 3031ms | 50th percentile (median) |
| P90 | 11018ms | 3557ms | 90th percentile |
Explore More
Compare more embeddings
See how all embedding models stack up. Compare OpenAI, Cohere, Jina AI, Voyage, and more. View comprehensive benchmarks, compare performance metrics, and find the perfect embedding for your RAG application.