OpenAI's open model. Smart for its size and fast because only part of it runs per word. Made by OpenAI.
gpt-oss 20B is a general chat model from OpenAI. OpenAI's open model. Smart for its size and fast because only part of it runs per word. Running it locally needs about 15.2 GB of memory, and the download is 14 GB. It is a mixture of experts model: only about 2.6 GB of its weights are read for each word, so it generates much faster than a dense model of the same size, but all 14 GB still has to fit in memory.
The least memory you need, by Apple chip:
Runs well on: RTX 3090, RTX 4060 Ti 16GB, RTX 4070 Ti Super, RTX 4080 / 4080 Super, RTX 4090, RTX 5060 Ti 16GB, RTX 5070 Ti, RTX 5080, RTX 5090, Radeon RX 7800 XT, Radeon RX 9070 XT, Radeon RX 7900 XTX. Runs slowly with help from system RAM on: RTX 3060 12GB, RTX 3070, RTX 3080 10GB, RTX 4060, RTX 4070 / 4070 Super, RTX 5060, RTX 5070.
As an Amazon Associate I earn from qualifying purchases.
About 15.2 GB of video memory or unified memory (the download is 14 GB, plus room for the conversation). On a Mac, the smallest configuration that runs it is 16 GB.
A 16 GB M4 Mac runs it just fits at about 13 words a second. An 8 GB M1 Mac cannot run it.
Yes, the model is free to download. Its licence is listed as Apache 2.0. That is a permissive licence that allows commercial use.
Install Ollama, then run: ollama run gpt-oss:20b
| Machine | Verdict | Speed |
|---|---|---|
| M1 8GB | Too big | |
| M4 16GB | Just fits | 13 words/s |
| M4 32GB | Runs great | 14 words/s |
| M4 Pro 48GB | Runs great | 32 words/s |
| M4 Max 128GB | Runs great | 65 words/s |
| M3 Ultra 512GB | Runs great | 98 words/s |
| RTX 4060 | Slow | 11 words/s |
| RTX 4070/4070 Super | Slow | 19 words/s |
| RTX 5070 Ti | Runs great | 92 words/s |
| RTX 3090 | Runs great | 96 words/s |
| RTX 5090 | Runs great | 183 words/s |
| Ryzen AI Max+ 395 128GB | Runs great | 26 words/s |
| CPU only 32GB | Slow | 7 words/s |