91 models
91 models
Hardware, cost and evidence come from each record. "Unmeasured" means we have not measured it on our own image yet.
Embedding · intfloat/multilingual-e5-large-instruct
Default text vector for search over transcripts, OCR text, descriptions, documents and web pages in 100+ languages; one space for query and passage text.
Embedding · multimodalembedding@001
Category-level text-to-video and text-to-image search in one 1408-d space (committed mAP 0.855 on a 10-video text eval).
Embedding · google/siglip-base-patch16-224
Text-to-image and image-to-image search over photos, products and video frames in one shared space.
Reranking · BAAI/bge-reranker-v2-m3
Re-ordering a short candidate list of text documents against a query (cross-encoder, multilingual), most useful where the dense ordering is weak.
Embedding · vidore/colqwen2-v1.0-merged
Text search over document page images (PDF pages, slides, infographics, tables) by late interaction, with no OCR step.
Embedding · gemini-embedding-2
One vector space for video segments, images, PDFs, audio and text queries; the default for object-level multimodal extractors (multimodal_extractor v2, universal_extractor).
Embedding · sentence-transformers/all-MiniLM-L6-v2
Small, fast English sentence vectors on CPU (22M params, 384 dims) for short text, dedup and clustering.
Classification · Falconsai/nsfw_image_detection
Fast binary normal/nsfw screening of images, and of video by the max score over sparse sampled frames.
Transcription · openai/whisper-large-v3-turbo
Multilingual speech-to-text on GPU at large-v3 quality with the decoder cut from 32 to 4 layers, so decoding is much faster than large-v3.
Generation · gemini-2.5-flash
The extractors' default for structured extraction and vision work: video description and OCR in multimodal_extractor v2, web page extraction, PDF layout (document_graph), scrolling-text OCR.
Embedding · facebook/dinov2-base
Image-to-image similarity on visual structure (layout, near-duplicates, the same object or scene) with no text in the loop.
Classification · eliasalbouzidi/distilbert-nsfw-text-classifier
Cheap binary safe/nsfw screening of English text such as captions, titles and transcripts.
Generation · gemini-3.5-flash-lite
The platform's default LLM (DEFAULT_GEMINI_MODEL) for cheap per-document work in retriever stages (enrich, filter, summarize) and for video-segment description and OCR in multimodal_extractor v1.
Face · insightface/buffalo_l
Face detection, 5-point alignment and 512-d face identity vectors for finding the same person across images and video frames.
Face · insightface/buffalo_sc
The fastest face detector option (SCRFD-500MF) for high-volume frames where small faces matter less.
Embedding · jinaai/jina-embeddings-v2-base-code
Code snippets and natural-language-to-code search across about 30 programming languages, with up to 8192-token inputs (ALiBi).
Embedding · lukewys/laion_clap
Sound-alike search over short audio windows (jingles, music beds, effects) and text-to-audio search ("upbeat jingle with bells") in one 512-d space.
Embedding · laion/CLIP-ViT-L-14-laion2B-s32B-b82K
Image-to-image similarity from a large web-trained ViT-L/14 image tower.
Embedding · google/siglip2-so400m-patch14-384
Logos, signage, brands and venue shots in concert footage (Live Nation visual arm); strongest SigLIP 2 below giant.
Generation · claude-sonnet-4-5-20250929
Multi-step tool use and reasoning; the default model of agent sessions (DEFAULT_AGENT_MODEL) and of agentic_enrich when no provider or model is given.
Detection · ustc-community/dfine-large-coco
Real-time DETR with fine-grained box refinement; among the highest COCO AP per latency of the permissive detectors.
Video and image Q&A · gemini-2.5-flash-lite
Cheapest Gemini video tagging and captioning at 1 fps.
Video and image Q&A · gemini-3-flash-preview
Stronger video reasoning than 2.5 Flash at Flash latency.
Directory pages for 422 public models. They are not in the catalog above, and Mixpeek does not run them as listed.