Can an RX 7600 XT run GPT OSS 120B?
Partly
Not entirely. About 7 of its 36 layers fit on the card and the rest run on the CPU, which brings generation down to roughly 11.8 tokens per second.
The numbers
GPT OSS 120B has 116.83 billion parameters across 36 layers, of which 5.7 billion are read for each token. The Radeon RX 7600 XT has 16 GB of VRAM at 288 GB/s, of which about 15.2 GB is left for a model once the operating system has taken its share.
| Quantisation | Total needed | Fits | Speed |
|---|---|---|---|
| Q8_0 | 116.39 GB | no | — |
| Q6_K | 90.01 GB | no | — |
| Q5_K_M | 78.17 GB | no | — |
| Q4_K_M | 66.48 GB | no | — |
| IQ4_XS | 58.59 GB | no | — |
| Q3_K_M | 53.97 GB | no | — |
| IQ3_XXS degraded | 42.4 GB | no | — |
| Q2_K degraded | 46.35 GB | no | — |
| IQ2_XXS degraded | 28.8 GB | no | — |
Nothing on this ladder fits, including the two-bit levels we do not recommend.
What to do instead
The straightforward answer is a smaller model from the same family. These do fit on a Radeon RX 7600 XT, at a quantisation worth using:
- GPT OSS 20B — 20.91B at Q5_K_M, 14.63 GB, about 74.7 tokens/s
You can also run it as is. Ollama and llama.cpp will put 7 of the 36 layers on the card and the rest on the CPU without being asked. It works, at roughly 11.8 tokens per second, which is fine for a batch job and tiring for a conversation.
Shortening the context window will not rescue this one: at 4k it still wants 66.39 GB against 66.48 GB at 8k, because the weights rather than the cache are what fill the card. The gap here is too large for settings to close.
Or the hardware that does run it
The cheapest device we track that runs GPT OSS 120B at a quantisation worth using is the Ryzen AI Max+ 395 96GB, around $1,999 for the whole machine, running it at Q4_K_M at about 51.8 tokens per second. Check the current price (affiliate link; indicative price reviewed 2026-09-01). Compare every option.
See how fast it feels
Radeon RX 7600 XT running GPT OSS 120B, 7 of 36 layers on the GPU
YouWhy does my model use more memory when the conversation gets longer?
Because of the KV cache. Every token you send leaves behind a key and a value vector in each layer of the model, and those stay in memory for as long as the conversation lasts.
The weights are a fixed cost: load a 4-bit 8B model and that is about 4.8 GB, whether you write one word or ten thousand. The cache is the part that grows, and it grows in a straight line with the number of tokens in the window.
How steeply depends on the model's attention design. With grouped-query attention, several query heads share one key-value pair, which cuts the cache by that ratio. Without it, every head keeps its own, and a long context can cost more memory than the weights themselves.
If you are short on memory, the first thing to try is lowering the context window in your runtime, not the quantisation.
Reading your question
Simulated from our estimate at a 512-token question, not a recording. Assumes nothing else is competing for the GPU.
Keep going
Everything the Radeon RX 7600 XT runs, the full breakdown for GPT OSS 120B, or check a different pairing in the calculator.