Written by Jakub Rusinowski · Last updated October 8, 2024
Yes
Yes — Aya Expanse 32B at Q4_K_M needs about 21.6 GB of the 24 GB usable on RTX 4090 Desktop (24 GB VRAM, 64 GB RAM) (~2.4 GB spare), at ~34.6 tok/s (estimated), with room for about 16,384 tokens of context.
Confidence: high · Recommended quantization: Q4_K_M · Estimated speed: ~34.6 tok/s
| Usable memory for models | 24 GB |
| Memory bandwidth | 1008 GB/s |
| Form factor | Desktop |
| Operating system | Windows or Linux |
| Memory upgradeable | Yes |
| Quant | Memory needed | Fits 24 GB? | Max context | Est. speed | Download |
|---|---|---|---|---|---|
| F16 | 66.7 GB | ✗ No | — | — | 64.6 GB |
| Q8_0 | 36.5 GB | ✗ No | — | — | 34.3 GB |
| Q6_K | 28.6 GB | ✗ No | — | — | 26.5 GB |
| Q5_K_M | 25 GB | ✗ No | — | — | 22.9 GB |
| Q4_K_M | 21.6 GB | ✓ Yes | 16K | ~34.6 tok/s | 19.5 GB |
| Q3_K_M | 15.9 GB | ✓ Yes | 32K | ~46.8 tok/s | 13.8 GB |
| Q2_K | 12.8 GB | ✓ Yes | 64K | ~58.2 tok/s | 10.6 GB |
| Variant | Needs at Q4_K_M | Fits? | Est. speed |
|---|---|---|---|
| Aya Expanse 32B | 21.6 GB | ✓ Fits | ~34.6 tok/s |
| Aya Expanse 8B | 6.7 GB | ✓ Fits | ~106.7 tok/s |
Ollama or llama.cpp (CUDA/ROCm) — vLLM if you need concurrent serving
Yes — Aya Expanse 32B at Q4_K_M needs about 21.6 GB of the 24 GB usable on RTX 4090 Desktop (24 GB VRAM, 64 GB RAM) (~2.4 GB spare), at ~34.6 tok/s (estimated), with room for about 16,384 tokens of context.
Q4_K_M — it needs about 21.6 GB of the 24 GB available, downloads as roughly 19.5 GB, and runs at an estimated 34.6 tokens/sec with up to 16K of context.
Nothing binding — the model fits with headroom and generates at a usable speed on this hardware.
Ollama or llama.cpp (CUDA/ROCm) — vLLM if you need concurrent serving
← Can I Run It? | Aya Expanse model page | Check your hardware