This page is part of the model documentation OpenSource Solutions keeps for the private AI assistants we build and host. We run open source models on our own servers and test them on real business questions before recommending them. All the models we test
AI model
Installed on our servers
gpt-oss:20b
OpenAI · released 2025-08-05
OpenAI's open-weight reasoning model: 21 billion parameters, of which 3.6 billion are active per token (mixture of experts). Built to run on a single 16 GB graphics card, with tool use and adjustable reasoning effort. OpenAI compares it to o3-mini on common benchmarks.
Completion Thinking Tools
Parameters
20.9B
MXFP4
Memory
13.8 GB
on the graphics card
Speed
95
tokens per second on our server
Max context
131072
tokens the model can read at once
Licence and cost
- License
- Apache 2.0
- Price
- Free, open weights. Runs on our servers: no per-token cost.
- Family
- gptoss
Which model for your assistant?
A model alone is not an assistant. OpenSource Solutions builds the harness around it: the connection to your documents and tools, the rules it follows, the hosting on your servers or ours, and the measurement on your own questions before you choose. Your data stays on infrastructure you control.