Local AI on Radeon RX 7900 XTX: which models run and how fast
Short answer
Radeon RX 7900 XTX has 24 GB of video memory, which holds 10 of Nightjar's eleven models entirely, the largest being Qwen 3.5 35B-A3B.
- Video memory
- 24 GB
- Models that fit fully
- 10 of 11
- Largest that fits
- Qwen 3.5 35B-A3B
- This is our test card
- no
Every model against this card
- Llama 3.2 1B — 0.75 GB, fits with 22.2 GB spare
- Qwen 3.5 2B — 1.2 GB, fits with 21.7 GB spare
- Qwen 3.5 4B — 2.6 GB, fits with 20.3 GB spare
- Gemma 4 E2B-it — 3.2 GB, fits with 19.7 GB spare
- Gemma 4 E4B-it — 5 GB, fits with 17.9 GB spare
- Qwen 3.5 9B — 5.3 GB, fits with 17.6 GB spare
- Gemma 4 12B-it — 6.6 GB, fits with 16.3 GB spare
- Gemma 4 26B-A4B-it — 15.8 GB, fits with 7.1 GB spare
- Qwen 3.5 27B — 15.6 GB, fits with 7.3 GB spare
- Qwen 3.5 35B-A3B — 20.5 GB, fits with 2.4 GB spare
- Gemma 4 31B-it — 30.4 GB, needs CPU offload
What the numbers here are and are not
The fit is arithmetic — a file size against a memory size, with about 1.1 GB added for context and runtime. The speeds elsewhere on this site were measured on RTX 3060 12 GB, Vulkan backend and nowhere else. Where a page says a figure was not measured, that is literal: we would rather leave a gap than publish an estimate that reads like a result.
Other ways people ask this
- best local llm for radeon rx 7900 xtx
- radeon rx 7900 xtx llm tokens per second
- I have a Radeon RX 7900 XTX. Which local model should I run for browser AI?