PerplexitySpecialist modelOpen weights
pplx-embed-v2-late-9b
Summary
Perplexity released pplx-embed-v2-late as open weights on Hugging Face on October 7, 2026. The 9B variant is the family's larger late-interaction retriever for text, images, and visual documents. It uses MaxSim instead of a single vector per document.
The model card lists 128 dimensions per token and 7.4 billion active parameters. This checkpoint shares its embedding space with the 0.6B variant, so different family sizes can encode documents and search queries.
Text and images must be batched separately. The open weights are licensed under MIT. On the public ViDoRe v3 set, Perplexity reports nDCG@10 of 65.2% for images and 64.7% for Markdown documents.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 2 published benchmark results
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 (Public, Image) (Show measurement, test conditions, and source)
- Source value
- 65.2
- Score
- 65.2
- Metric
- nDCG@10
- Unit
- %
- Benchmark version
- 3
- Category
- retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Perplexity pplx-embed-v2-late-9b model card
- Status
- active
- Retrieved at
- 2026-11-07
- Methodology
- Vendor-reported result. Public ViDoRe v3 set, documents represented as images.
ViDoRe v3 (Public, Markdown) (Show measurement, test conditions, and source)
- Source value
- 64.7
- Score
- 64.7
- Metric
- nDCG@10
- Unit
- %
- Benchmark version
- 3
- Category
- retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Perplexity pplx-embed-v2-late-9b model card
- Status
- active
- Retrieved at
- 2026-11-07
- Methodology
- Vendor-reported result. Public ViDoRe v3 set, documents represented as Markdown.
Similar models
More models of the same kind as pplx-embed-v2-late-9b, chosen by model line, developer, size, price and release date.
- pplx-embed-v2-context-9bPerplexity · same developer
- pplx-embed-v2-context-9b-previewPerplexity · same developer
- pplx-embed-v2-late-0.6bPerplexity · same developer
- pplx-embed-v2-context-0.6bPerplexity · same developer
- AstaBrief-8BAllen Institute for AI · similar size
- FLUX 3 ActionBlack Forest Labs · similar size
- Qwen-Drive-1.0-4BQwen · similar size
- Unlimited OCR 2Baidu · similar size
All specifications
Architecture
Input and output
- Modalities
- Text, images, and visual documents. Mixed text and image batches are not supported.
- Source: Modalities
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
- Shutdown
- None announced
- no source
Further details
- Embedding output
- One 128-dimensional vector per token, scored with MaxSim. The 0.6B and 9B models share an embedding space.
- Source: Embedding output
Sources and data date
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.