gpt-oss 20B

Good fit for

People who want to evaluate OpenAI open weights locally rather than use the separate ChatGPT service.

Why consider it

The exact runtime entry documents configurable reasoning effort, function calling, MXFP4 packaging and an Apache 2.0 license.

Main limitations

Tool-call capability is not a bundled browser or Python executor, so Web Search and Code Execution remain unassigned. The catalog says as little as 16GB memory for this format, but actual memory needs grow with context and runtime overhead. It is not equivalent to a ChatGPT subscription.

Cost and access

Local package without a hosted subscription or per-token inference charge. Hardware, electricity and setup remain separate. License: Apache 2.0.

Getting started

Install a compatible Ollama version, review the model license, then run: ollama run gpt-oss:20b. API access means the local runtime endpoint, not an included hosted API.