Curiosity · AI Model
Reka Vision
Reka Vision is Reka AI's enterprise multimodal product — a platform (rather than a single model) that wraps Reka's Core and Flash family to provide image, video, and document understanding at scale. It offers visual search, captioning, chapter generation, and Q&A over media libraries, aimed at media, e-commerce, and surveillance customers.
Model specs
- Vendor
- Reka AI
- Family
- Reka
- Released
- 2024-09
- Context window
- 128,000 tokens
- Modalities
- text, vision, video
- Input price
- n/a
- Output price
- n/a
- Pricing as of
- 2026-04-20
Strengths
- Built on Reka's frontier multimodal stack
- Supports long-form video understanding
- Enterprise deployment options (VPC, on-prem)
Limitations
- Pricing is sales-led — no public per-token rate
- Smaller ecosystem than hyperscaler visual AI products
- Requires careful pipeline integration for best accuracy
Use cases
- Enterprise video search and chapter generation
- E-commerce visual product search
- Content moderation pipelines with context awareness
- Media intelligence and surveillance review
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| Perception Test | ~60% | 2026-04 |
| MMMU | ~56% | 2026-04 |
| VideoMME | ~55% | 2026-04 |
Frequently asked questions
What is Reka Vision?
Reka Vision is Reka AI's enterprise product for multimodal understanding of images and video. It combines Reka Core/Flash models with retrieval, search, and captioning workflows.
How is Reka Vision different from Reka Core?
Reka Core is a model; Reka Vision is a product that uses Reka's models (Core and Flash) plus retrieval and indexing infrastructure to operate on large video and image libraries.
Sources
- Reka Vision homepage — accessed 2026-04-20
- Reka AI research page — accessed 2026-04-20