Model overview
gpt-oss-20b is described on the OpenAI page of NVIDIA Build as a smaller Mixture of Experts (MoE) text-only LLM. Its stated purpose is efficient AI reasoning and math. It is one of two OpenAI models listed on that page, the other being whisper-large-v3 for speech recognition.
- Smaller Mixture of Experts (MoE) architecture
- Text-only large language model
- Published under OpenAI on NVIDIA Build
Capabilities and tags
The listing categorises gpt-oss-20b with tags that indicate its intended tasks. These cover text-to-text generation, chat, reasoning and math, which aligns with readers searching for a gpt-oss-20b reasoning model.
- text-to-text
- chat
- reasoning
- math
Availability and usage
The model is marked as downloadable and is offered with a free endpoint on NVIDIA Build, where models can be tried through NVIDIA NIM APIs. The listing records 19M API calls in the last 30 days. The entry was last updated on 5 August 2025.
- Downloadable
- Free endpoint
- 19M API calls in the last 30 days
- Last updated 5 August 2025
Sources
Last verified 02 Oct 2026. This entry is compiled from the public web pages listed above. Nothing here is stated that those pages do not, and each of them was read on the date shown.