OpenAI · Reasoning Models

gpt-oss-20b

Name: gpt-oss-20b
Author: OpenAI

OpenAI's smaller open-weight reasoning model: a 21B mixture-of-experts with 3.6B active parameters that runs on machines with 16 GB of memory.

Official page

Run it locally

With Ollama installed, download and start the default version with:

ollama run gpt-oss:20b

Variant	Download
gpt-oss:20b	14GB

Memory: About 14GB for the default gpt-oss:20b version (4-bit)
Formats: Safetensors (Hugging Face), GGUF (Ollama, llama.cpp, LM Studio)

The download size is roughly the memory (RAM or GPU VRAM) needed to load a variant; allow extra for long contexts. See our hardware guide.

Overview

gpt-oss-20b is one of two open-weight models OpenAI released in August 2025 under Apache 2.0, its first open-weight language models since GPT-2. It supports adjustable reasoning effort and tool use.

Highlights

21B total, 3.6B active parameters
Adjustable reasoning effort (low, medium, high)
128K-token context window
Apache 2.0 license

Typical uses

Local reasoning assistants, agents with tool use and on-premises deployments.

Things to consider

Text only.
Uses the “harmony” response format, which tools must support.
Download sizes refer to the default 4-bit quantized versions in the Ollama library; full-precision weights are larger.

Specifications are taken from the model card on Hugging Face, the Ollama library and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.

#Reasoning Models #Open Weights #Local AI #OpenAI