TinyLlama on AMD Radeon RX 7900 XTX · 24GB VRAM
✅ Runs comfortably — TinyLlama 1.1B @ FP16
TinyLlama 1.1B at FP16 needs ~2.7 GB (weights 2.0 GB + KV 176 MB + overhead 512 MB @ 8K ctx) of your 21.6 GB usable VRAM — plenty of headroom. Expect ~170–280 tok/s (very fast).
context @ FP16: comfortable to 2K
| quant | needs | speed |
|---|---|---|
| ✓ FP16 | 2.7 GB | ~170–280 tok/s (very fast) |
| ✓ Q8_0 | 1.8 GB | ~290–490 tok/s (very fast) |
| ✓ Q4_K_M | 1.3 GB | ~470–780 tok/s (very fast) |
$runlocal install tinyllama:1.1b
for your own README — links back here
Same model, other machines: Apple M1 · Apple M1 Pro · Apple M2 · Apple M3 Pro · Apple M2 Pro · Apple M4 Pro
Also runs on a AMD Radeon RX 7900 XTX: DeepSeek-R1 (distill) · Qwen3 · Qwen2.5 · Gemma 3 · Mistral Small 3
Best models for a AMD Radeon RX 7900 XTX · 24GB VRAM →
Check any combo yourself: open the checker →