Specialist modelActive
Cohere Embed v5 Pro
Cohere
- Released
- September 30, 2026
- Data date
- October 7, 2026
Cohere Embed v5 Pro is optimized for retrieval quality. Cohere reports improvements over Embed 4 on visually rich documents, financial filings, PDFs, code, and multilingual retrieval.
Pro shares its embedding space with Embed v5 Fast, allowing a corpus indexed with Pro to be searched using Fast queries. Both variants accept text, images, and mixed text-image input.
Specifications and access
| Specification | Value and source |
|---|---|
| API model ID | embed-v5.0-proSource |
| Input modalities | Text, images, and mixed text-image inputs, such as PDF pagesSource |
| Output modalities | Embedding vectors with 256, 512, 768, 1,024, 1,536, or 2,048 dimensionsSource |
| Focus | Highest retrieval quality, especially for offline indexing and quality-critical searchSource |
| Languages | More than 100 languagesSource |
| Context window | 128K tokens per Cohere; no exact integer count publishedSource |
| Access | Cohere Embed API, Microsoft Foundry, Amazon SageMaker, or Model VaultSource |
Page 1 of 4
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
ViDoRe V3, eight-domain mean (Show measurement, test conditions, and source)
- Source value
- 85.8
- Score
- 85.8
- Metric
- RCP-nDCG@10
- Unit
- points
- Benchmark version
- ViDoRe V3
- Category
- visual document retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Eight visually rich document domains; ViDoRe-authored curated parsed-text outputs.
- Context
- The announcement does not specify output dimension or vector precision for this chart.
FinanceBench retrieval (Show measurement, test conditions, and source)
- Source value
- 80.1
- Score
- 80.1
- Metric
- unknown
- Unit
- No unit provided
- Benchmark version
- FinanceBench
- Category
- financial retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. The launch post reports a scalar benchmark score but does not identify the per-benchmark metric, score unit, dataset split, output dimension, or precision.
- Context
- The post labels this value a score but does not publish the metric, unit, or split for this benchmark row.
FinQA retrieval (Show measurement, test conditions, and source)
- Source value
- 90
- Score
- 90
- Metric
- unknown
- Unit
- No unit provided
- Benchmark version
- FinQA
- Category
- financial retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. The launch post reports a scalar benchmark score but does not identify the per-benchmark metric, score unit, dataset split, output dimension, or precision.
- Context
- The post labels this value a score but does not publish the metric, unit, or split for this benchmark row.
ViDoRe V3 Finance retrieval (Show measurement, test conditions, and source)
- Source value
- 85
- Score
- 85
- Metric
- unknown
- Unit
- No unit provided
- Benchmark version
- ViDoRe V3 Finance
- Category
- financial retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. The launch post reports a scalar benchmark score but does not identify the per-benchmark metric, score unit, dataset split, output dimension, or precision.
- Context
- The post labels this value a score but does not publish the metric, unit, or split for this benchmark row.
Parsed PDF retrieval, full-suite mean (Show measurement, test conditions, and source)
- Source value
- 84.8
- Score
- 84.8
- Metric
- RCP-nDCG@10
- Unit
- points
- Benchmark version
- Cohere parsed-PDF suite
- Category
- parsed-document retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Parsed-document suite; PDF content was parsed using Gemini 1.5 Flash. The plotted public-benchmark rows are a subset; the full-suite average also includes unnamed additional datasets.
- Context
- The exact output dimension and precision are not stated for this evaluation.
Parsed PDF retrieval, RepairBench (Show measurement, test conditions, and source)
- Source value
- 86.9
- Score
- 86.9
- Metric
- RCP-nDCG@10
- Unit
- points
- Benchmark version
- Cohere parsed-PDF suite
- Category
- parsed-document retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Parsed-document suite; PDF content was parsed using Gemini 1.5 Flash. The plotted public-benchmark rows are a subset; the full-suite average also includes unnamed additional datasets.
- Context
- The exact output dimension and precision are not stated for this evaluation.
Parsed PDF retrieval, CoFiF (FR) (Show measurement, test conditions, and source)
- Source value
- 87.4
- Score
- 87.4
- Metric
- RCP-nDCG@10
- Unit
- points
- Benchmark version
- Cohere parsed-PDF suite
- Category
- parsed-document retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Parsed-document suite; PDF content was parsed using Gemini 1.5 Flash. The plotted public-benchmark rows are a subset; the full-suite average also includes unnamed additional datasets.
- Context
- The exact output dimension and precision are not stated for this evaluation.
Parsed PDF retrieval, FinanceBench (Show measurement, test conditions, and source)
- Source value
- 81.5
- Score
- 81.5
- Metric
- RCP-nDCG@10
- Unit
- points
- Benchmark version
- Cohere parsed-PDF suite
- Category
- parsed-document retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Parsed-document suite; PDF content was parsed using Gemini 1.5 Flash. The plotted public-benchmark rows are a subset; the full-suite average also includes unnamed additional datasets.
- Context
- The exact output dimension and precision are not stated for this evaluation.
Parsed PDF retrieval, MPQMA (Show measurement, test conditions, and source)
- Source value
- 82.3
- Score
- 82.3
- Metric
- RCP-nDCG@10
- Unit
- points
- Benchmark version
- Cohere parsed-PDF suite
- Category
- parsed-document retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Parsed-document suite; PDF content was parsed using Gemini 1.5 Flash. The plotted public-benchmark rows are a subset; the full-suite average also includes unnamed additional datasets.
- Context
- The exact output dimension and precision are not stated for this evaluation.
Parsed PDF retrieval, PolicyQA (Show measurement, test conditions, and source)
- Source value
- 80.4
- Score
- 80.4
- Metric
- RCP-nDCG@10
- Unit
- points
- Benchmark version
- Cohere parsed-PDF suite
- Category
- parsed-document retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Parsed-document suite; PDF content was parsed using Gemini 1.5 Flash. The plotted public-benchmark rows are a subset; the full-suite average also includes unnamed additional datasets.
- Context
- The exact output dimension and precision are not stated for this evaluation.
Fused text-image retrieval, five-dataset mean (Show measurement, test conditions, and source)
- Source value
- 82.3
- Score
- 82.3
- Metric
- RCP-nDCG@10 (chart heading); nearby caption says nDCG@10
- Unit
- points
- Benchmark version
- Cohere fused text-image suite
- Category
- fused text-image retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Fused text-image retrieval with text queries. Five datasets in the reported suite.
- Context
- The chart says RCP-nDCG@10, while its nearby caption says nDCG@10. The source metric label is inconsistent.
Fused text-image retrieval, High Finance (Show measurement, test conditions, and source)
- Source value
- 94.8
- Score
- 94.8
- Metric
- RCP-nDCG@10 (chart heading); nearby caption says nDCG@10
- Unit
- points
- Benchmark version
- Cohere fused text-image suite
- Category
- fused text-image retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Embed 5 release and evaluation
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Cohere API variant embed-v5.0-pro. The release does not specify a particular output dimension or vector precision for this score. Fused text-image retrieval with text queries. The fused vector combines page image and document metadata; RepairBench averages multiple multilingual query subsets; High Finance is an internal Cohere-annotated set.
- Context
- The chart says RCP-nDCG@10, while its nearby caption says nDCG@10. The source metric label is inconsistent.
Page 1 of 4
Sources and data date
| Type | Evidence and data date |
|---|---|
| Research status | Research date October 7, 2026. 2 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | full linked evaluation workbook and per-row configurations Original wording Cohere links a full-results spreadsheet, but it was inaccessible during this read. The public post omits output dimension, vector precision, sample counts, and in several tables the exact metric mapping. |
| Unresolved | fused and page-image metric label Original wording The chart headings call these results RCP-nDCG@10, while the nearby prose captions call them nDCG@10. Both labels are retained on each row rather than resolved by assumption. |
| Additional source | Cohere Embed 5 release notes (retrieved October 7, 2026) · Cohere Embed 5 release notes · Editorial description reviewed October 7, 2026 |
| Additional source | Cohere Embed 5 release and evaluation (retrieved October 7, 2026) |