WizardCoder-15.5B

Training compute

1.1×10²³ FLOP

Parameters

15.5B

Published

Jun 14, 2023

WizardCoder-15.5B is an AI model developed by Microsoft (United States), first published in June 2023. It works in the language domain, on tasks such as code generation.

Training it took an estimated 1.1×10²³ FLOP of compute (estimation method: operation counting). The model has 15,500,000,000 parameters. It was trained on roughly 209.7M datapoints.

Access: Open weights (restricted use). Its weights are openly available. It is built on top of StarCoder. The reference paper has 952 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: Microsoft
Country of organization: United States
Domain: Language
Task: Code generation
Training compute: 1.1×10²³ FLOP
Compute estimation method: Operation counting
Parameters: 15,500,000,000
Dataset size: 209.7M
Model accessibility: Open weights (restricted use)
Open weights: Yes
Base model: StarCoder
Citations: 952
Epoch confidence: Likely