DeepSeek-Coder-V2 236B

Training compute

1.3×10²⁴ FLOP

Parameters

236B

Published

Jun 17, 2024

DeepSeek-Coder-V2 236B is an AI model developed by DeepSeek (China), first published in June 2024. It works in the language domain, on tasks such as code generation and code autocompletion.

Training it took an estimated 1.3×10²⁴ FLOP of compute (estimation method: operation counting). The model has 236,000,000,000 parameters. It was trained on roughly 3.2T datapoints.

Access: Open weights (restricted use). Its weights are openly available. It is built on top of DeepSeek-V2 (MoE-236B). Epoch AI rates the confidence of this record as confident.

Full record

Organization: DeepSeek
Country of organization: China
Domain: Language
Task: Code generation, Code autocompletion
Training compute: 1.3×10²⁴ FLOP
Compute estimation method: Operation counting
Parameters: 236,000,000,000
Dataset size: 3.2T
Model accessibility: Open weights (restricted use)
Open weights: Yes
Base model: DeepSeek-V2 (MoE-236B)
Epoch confidence: Confident