Megatron MarketsWholesale
Sign inContact Us
Models Model details

OpenAI

OpenAI: gpt-oss-120b

openai/gpt-oss-120b

Enquire about this model Request a quote

gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases. It activates 5.1B parameters per forward pass and is optimized to run on a single H100 GPU with native MXFP4 quantization. The model supports configurable reasoning depth, full chain-of-thought access, and native tool use, including function calling, browsing, and structured output generation.

Input
$0.15

/ 1M tokens

Output
$0.6

/ 1M tokens

Context window
131.07K

tokens

Limits

Context

131,072

tokens

Max output

65,536

50% of the context window

Parameters

116,829,156,672

Model facts

Author

OpenAI

Tokenizer

GPT

Modality

Text -> Text

License

Apache-2.0

Knowledge cutoff

2024-06-30

Created

Aug 5, 2025

Hugging Face

openai/gpt-oss-120b

Reasoning

RequiredDefault MediumHighMediumLow

Publication status

Hugging Face: verifiedLicense: publishedParameter count: published

Modalities

Input · TextOutput · Text

Generation controls

Frequency penaltyInclude reasoningLogit biasLogprobsMax tokensMin pPresence penaltyReasoningReasoning effortRepetition penaltyResponse formatSeedStopStructured outputsTemperatureTool choiceToolsTop aTop kTop logprobsTop p

Capability scores

3.7
Agentic index

Better than 26% of models compared

30.4
Coding index

Better than 25% of models compared

11.6
Intelligence index

Better than 28% of models compared

Creative tasks

Category performance

Data visualization
Top 93%

987 Elo

43.6% win

Games
Top 90%

995 Elo

40.5% win

UI components
Top 95%

926 Elo

35.7% win

Code
Top 95%

965 Elo

33.4% win

Websites
Top 94%

969 Elo

32.5% win

3D
Top 94%

902 Elo

29.4% win