Model
Inference & Serving
Reka Flash 3.1
1Efficient 2multimodal 3language 4model 5for 6enterprises
the six Ws · specification
W1
Who
Reka AI develops the Reka Flash line of efficient multimodal language models.
W2
What
Reka Flash 3.1 is a compact multimodal model handling text, image, and reasoning tasks with fast inference.
W3
Where
Accessible through the Reka AI platform and API, with select checkpoints on Hugging Face.
W4
When
Reka Flash launched April 2024, with the 3.1 update released in 2025.
W5
Why
Designed to deliver near-frontier multimodal quality at low latency and cost for production deployments.
W6
With
Runs via Reka's hosted API or self-hosted inference using released weights on Hugging Face.
W7
Watch
Weights partially open (earlier Reka Flash releases under Apache 2.0), hosted API requires a Reka API key; maintained by Reka AI.
alternatives
multimodalefficient-LLMvision-languagemultilingualedge-deployable