Reka

Reka is a multimodal AI platform from Reka AI that natively understands video, audio, images, and text. It processes real-world media as first-class inputs rather than bolting vision onto a text model. Reka is open-source, with select model weights released under permissive licences, and supports self-hosting in cloud, on-premise, VPC, or fully air-gapped environments.

Enterprise teams in defence, healthcare, and finance use Reka because it handles sensitive media without sending data to third-party APIs. The Reka Playground is free for browser-based testing, and Reka Clip is a free creator tool for searching long videos by question.

Features:

• •Native Video Understanding Reka processes video with both audio and visual context together, so you search long recordings, detect scenes, and generate captions without separate transcription or frame extraction.
• •Audio Intelligence Transcribe audio, identify speakers, and extract key insights from meetings, podcasts, and call recordings using a pipeline that runs natively alongside the video and image models.
• •Image Analysis Describe images in detail, answer visual questions, and generate multimodal embeddings for search and classification across mixed media types.
• •Open-Source Model Weights Select Reka models are released open-source under permissive licences, letting research teams inspect architecture, fine-tune on private data, and contribute back without vendor lock-in.
• •Self-Hosting And Air-Gapped Deployment Run Reka in your own cloud, on-premise servers, VPC, or fully air-gapped data centres, covering regulated industries that cannot send data to external APIs.
• •Free Reka Playground Test all core models in the browser at no cost. Upload video, audio, or image files and see results before committing to a paid plan or self-hosted setup.
• •Reka Clip For Creators A free tool that lets creators search long videos by asking questions, chat with video content, and auto-generate highlight clips from uploaded footage.
• •Multimodal Embeddings Generate embeddings from video, audio, image, or text inputs, then use them for semantic search, clustering, or recommendation systems across mixed media.
• •Unified Text Capabilities Reka also handles standard text tasks like summarisation, Q&A, and reasoning, sharing the same foundation as the multimodal models.

Pricing:

• •Free Playground $0/mo, test all core models in the browser with sample or uploaded files.
• •Reka Clip $0/mo, free creator tool for video search and highlight generation.
• •API Access Pay-per-use pricing based on tokens processed; contact Reka AI directly for current per-token rates.
• •Enterprise Self-Hosting Custom pricing per year for on-premise, VPC, or air-gapped deployment with SLAs and dedicated support.

Limitations:

• •Opaque API pricing Reka does not publish per-token rates on its public pricing page. You must contact sales for a quote, which makes cost comparison with Gemini or GPT-4o difficult before committing.
• •No mobile application Reka runs as a web platform only. There is no iOS or Android app, so on-the-go media processing requires a mobile browser.
• •Limited third-party integrations Unlike ChatGPT with its plugin ecosystem or Gemini with Google Workspace, Reka does not offer a broad marketplace of pre-built connectors or extensions.
• •Playground feature limits The free Playground is designed for evaluation, not production workloads. File-size limits and request quotas constrain heavy testing sessions.
• •Smaller community than mainstream LLMs Reka has fewer Reddit threads, YouTube tutorials, and Stack Overflow answers than Claude or ChatGPT, so self-service troubleshooting takes longer.