Reka Core

Training compute

8.4×10²⁴ FLOP

Parameters

67B

Published

Apr 15, 2024

Reka Core is an AI model developed by Reka AI (United States), first published in April 2024. It works in the multimodal, language, vision and 2 more domain, on tasks such as chat, language modeling/generation, image captioning and 8 more.

Training it took an estimated 8.4×10²⁴ FLOP of compute (estimation method: hardware). The model has 67,000,000,000 parameters. Training ran on NVIDIA A100,NVIDIA H100 SXM5 80GB.

Access: API access. Its weights are not openly released. Epoch AI rates the confidence of this record as speculative.

Full record

Organization: Reka AI
Country of organization: United States
Domain: Multimodal, Language, Vision, Video, Speech
Task: Chat, Language modeling/generation, Image captioning, Code generation, Code autocompletion, Question answering, Visual question answering, Video description, Speech recognition (ASR), Speech-to-text, Quantitative reasoning
Training compute: 8.4×10²⁴ FLOP
Compute estimation method: Hardware
Parameters: 67,000,000,000
Training hardware: NVIDIA A100, NVIDIA H100 SXM5 80GB
Numerical format: BF16
Model accessibility: API access
Open weights: No
Epoch confidence: Speculative