gpt-oss:120b
An open-source 120B-parameter large language model from OpenAI that uses a MoE architecture. It can run on consumer-grade hardware through quantization, making it suitable for local inference and orchestration tasks.
Timeline (last 90 days)
The 77GB coordinator model gpt-oss:120b cannot share the same machine with the worker models it calls, and each task delegation triggers about 100GB of model swapping, with a simple greeting taking nearly five minutes.
After migrating the worker models to a second host with a 5600G equipped with a GTX 1060, gpt-oss:120b stayed resident throughout a long run of 1 hour and 46 minutes with no swapping at all, and orchestration reliability surpassed Gemma.
When the 120B coordinator is resident (about 77GB), only about 3GB of memory is free in the system; in a multi-image run the first image completes and the remaining images time out at the 300-second limit and fall back to placeholders.
The only real advantage of 120B over 20B is more reliable orchestration, but even this did not solve the deeper reliability problem it was supposed to solve.
All Facts (4)
The only real advantage of 120B over 20B is more reliable orchestration, but even this did not solve the deeper reliability problem it was supposed to solve.
50%UnverifiedThe 77GB coordinator model gpt-oss:120b cannot share the same machine with the worker models it calls, and each task delegation triggers about 100GB of model swapping, with a simple greeting taking nearly five minutes.
50%UnverifiedAfter migrating the worker models to a second host with a 5600G equipped with a GTX 1060, gpt-oss:120b stayed resident throughout a long run of 1 hour and 46 minutes with no swapping at all, and orchestration reliability surpassed Gemma.
50%UnverifiedWhen the 120B coordinator is resident (about 77GB), only about 3GB of memory is free in the system; in a multi-image run the first image completes and the remaining images time out at the 300-second limit and fall back to placeholders.
50%