TwelveLabs
Video understanding models that search and analyse what happens in video
Overview
TwelveLabs builds video foundation models — Marengo for multimodal embeddings and search, Pegasus for analysis and structured metadata — sold via API and Amazon Bedrock, plus the Jockey video agent and Rodeo creator app. Free plan with 10 hours; pay-as-you-go indexing $2.50 per video hour; Enterprise custom.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 2 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 10 hours of indexing and analysis
- 90-day index access
- 100 videos per index
- Search, Analyze and Embed
- Indexing $2.50 per hour
- Search $4 per 1,000 queries
- Analyze $1.75/hour input + $7.50 per 1M output tokens
- Marengo 3.5 embeddings per 1M tokens
- Custom limits
- Fine-tuning
- Private or self-managed deployment
What you can produce with TwelveLabs
- Marengo multimodal video embeddings
- Video search
- Pegasus video analysis and structured metadata
- Jockey video agent (research preview)
- Amazon Bedrock availability
- Python and JavaScript SDKs
TwelveLabs is one of the most capable specialists in video understanding, with strong search and analysis models, clear usage pricing and fresh funding behind a move into agents and apps. Models change quickly — Pegasus 1.2 has already been retired from its API — its benchmark claims are its own, and independent reviews are non-existent. A strong choice for teams with large video archives; pin model versions in production.
What we found
TwelveLabs builds foundation models that understand video the way people do — what happens, who is speaking, what is shown and what is said — so that large video libraries can be searched, summarised and analysed. Its Marengo models create multimodal embeddings across video, audio, images and documents, with Marengo 3.5 released in August 2026 for embeddings and Marengo 3.0 powering search, while Pegasus 1.5, released in April, generates text and extracts timestamped, schema-structured metadata from videos up to two hours long through one API call. Pegasus 1.2 was retired from its own API in August, although Amazon Bedrock still offers it. Beyond the API, TwelveLabs now offers Jockey, a video agent in research preview available through Claude's connectors directory, and Rodeo, an app for creators. TwelveLabs reports its models outperforming rivals on its own benchmarks, and its Python and JavaScript SDKs are public but carry no open-source licence.
Twelve Labs Inc. was founded in 2021, is headquartered in San Francisco with operations in Seoul and Pangyo and new offices in New York and London, and is led by co-founder and chief executive Jae Lee. In July 2026 it raised a $100 million Series B co-led by NEA and NAVER Ventures, with Amazon among the investors, bringing total funding to about $200 million according to Korean press. Its models are available on Amazon Bedrock, and named users include the UNICEF Korea archive and CBS.
The free plan offers ten hours shared across indexing and analysis, 90-day index retention and up to 100 videos per index, with no card. Developer pricing is pay-as-you-go: indexing costs $2.50 per video hour, search $4 per 1,000 queries, analysis $1.75 per hour of input plus $7.50 per million output tokens, and embeddings are priced per million tokens by media type. Enterprise contracts add custom limits, fine-tuning and private or self-managed deployment.
Independent reviews are essentially absent; G2 lists no reviews. TwelveLabs suits media, sports and security teams that need to search and analyse large video archives.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including TwelveLabs's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with TwelveLabs unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to TwelveLabs directly →
Spotted something out of date? Suggest an update →
Alternatives to TwelveLabs
More in Video & Audio
