Reka AI'sOctober 2024 Flash update expanded the model's video understanding to include speech and environmental sounds alongside visual content. The release described processing three to five minutes of video at a time, compared with up to one minute in the first version.
That combination supports questions about what happens in a clip using both what is visible and what is audible. The release also explicitly says this version could not locate events at exact timestamps. Its understanding of a video's sequence should therefore be distinguished from the ability to identify a precise moment for an editor or search system.
Terms in this piece · Glossary
structured output — Forcing a model's response to match a schema, so downstream code can parse it instead of guessing at prose.
multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
