Pangu Ultra

Training compute

1.1×10²⁵ FLOP

Parameters

135B

Published

Apr 10, 2025

Pangu Ultra is an AI model developed by Huawei (China), first published in April 2025. It works in the language domain, on tasks such as code generation and language modeling/generation.

Training it took an estimated 1.1×10²⁵ FLOP of compute (estimation method: operation counting,comparison with other models). The model has 135,000,000,000 parameters. It was trained on roughly 13.2T datapoints. Training ran on 8,192 Huawei Ascend 910B.

Access: Hosted access (no API). Its weights are not openly released. The reference paper has 16 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Huawei
Country of organization: China
Domain: Language
Task: Code generation, Language modeling/generation
Training compute: 1.1×10²⁵ FLOP
Compute estimation method: Operation counting, Comparison with other models
Parameters: 135,000,000,000
Dataset size: 13.2T
Training hardware: Huawei Ascend 910B
Chips used: 8,192
Training power draw: 6.4 MW
Model accessibility: Hosted access (no API)
Open weights: No
Citations: 16
Epoch confidence: Confident