Reka

Reka is a multimodal AI model builder for enterprises that need production deployments of models and agentic platforms that can reason over text, images, video, and audio.

Key capabilities include:

• Multimodal models trained from scratch (Core, Flash, Edge, plus Spark) designed for different latency, compute, and task complexity needs
• Vision Platform for unified video and image understanding, including captioning, detection, embeddings, Q&A, and search
• Speech for audio understanding beyond transcription, focused on extracting meaning, context, and insights
• Developer APIs and docs with REST endpoints plus SDKs (including Python and JavaScript), tutorials, and sample apps

Pricing: Unknown (demo and enterprise contact available).

What makes it notable: it pairs an enterprise deployment focus (including cloud, on-prem, VPC, and air-gapped options) with a track record of open-sourcing components such as reasoning models, evaluation tools, and a quantization stack, alongside published technical research on its model family.