gpt-oss-120b

Training compute

4.9×10²⁴ FLOP

Parameters

116.8B

Published

Aug 5, 2025

gpt-oss-120b is an AI model developed by OpenAI (United States), first published in August 2025. It works in the language domain, on tasks such as language modeling/generation and question answering.

Training it took an estimated 4.9×10²⁴ FLOP of compute (estimation method: hardware). The model has 116,830,000,000 parameters. Training ran on NVIDIA H100 SXM5 80GB.

Access: Open weights (unrestricted). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: OpenAI
Country of organization: United States
Domain: Language
Task: Language modeling/generation, Question answering
Training compute: 4.9×10²⁴ FLOP
Compute estimation method: Hardware
Parameters: 116,830,000,000
Training hardware: NVIDIA H100 SXM5 80GB
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Epoch confidence: Confident

Benchmark results

01	ALE Bench	Performance	575.6
02	Epoch Capabilities Index	ECI Score	140.6
03	Lech Mazur Writing	Mean score	7.71
04	METR Time Horizons	average_score	0.57
05	Surface Evolver Bench	Mean score	0.25
06	SimpleBench	Score (AVG@5)	0.22
07	Terminal-Bench	Accuracy mean	0.19
08	Vending-Bench 2	Score	-21.53