Can I Run Yi 1.5 Family on 8 GB system RAM?

Newer alternative. 01.AI has not released a newer model in this library. If you are choosing today, Qwen 3 is a comparable current option. View Qwen 3 →

Written by Jakub Rusinowski · Last updated May 13, 2024

Yes

Yes — Yi 1.5 9B Chat at Q3_K_M needs about 5.4 GB of the 6.4 GB usable on 8 GB system RAM (~1 GB spare), at ~15.6 tok/s (estimated), with room for about 16,384 tokens of context.

Confidence: high · Recommended quantization: Q3_K_M · Estimated speed: ~15.6 tok/s

8 GB system RAM — what it gives a model

Usable memory for models	6.4 GB
Memory bandwidth	90 GB/s

Yi 1.5 Family on 8 GB system RAM: memory by quantization

Quant	Memory needed	Fits 6.4 GB?	Max context	Est. speed	Download
F16	19.3 GB	✗ No	—	—	17.7 GB
Q8_0	11 GB	✗ No	—	—	9.4 GB
Q6_K	8.8 GB	✗ No	—	—	7.2 GB
Q5_K_M	7.9 GB	✗ No	—	—	6.3 GB
Q4_K_M	6.9 GB	✗ No	—	—	5.3 GB
Q3_K_M	5.4 GB	✓ Yes	16K	~15.6 tok/s	3.8 GB
Q2_K	4.5 GB	✓ Yes	16K	~19.4 tok/s	2.9 GB

Which Yi 1.5 Family sizes fit

Variant	Needs at Q4_K_M	Fits?	Est. speed
Yi 1.5 34B Chat	23.6 GB	✗ Too large	—
Yi 1.5 9B Chat	6.9 GB	✗ Too large	—

What to watch out for

Only ~1 GB of headroom at Q3_K_M: a longer context or a second application can push this into swapping.
Q3_K_M is the only quantization that fits, and it is a heavily degraded one — expect noticeably worse output than the same model at Q4_K_M. A smaller model at Q4 is usually the better trade.
2 larger variants of Yi 1.5 Family do not fit and would need CPU offload or different hardware.
These figures assume CPU-only inference. Any discrete GPU, even an 8 GB one, will be several times faster for models that fit in its VRAM.

Recommended setup

llama.cpp (CPU build) or Ollama — both run without a GPU

How these numbers are calculated

Memory need = quantized weights + KV cache at 8,192 tokens (f16) + 0.8 GB runtime overhead.
6.4 GB of the 8 GB is treated as usable for model weights (80% — the rest is the OS and running applications).
DDR5-5600 dual channel at 89.6 GB/s peak. CPU decode is assumed to sustain 35% of that peak, because CPU inference is not purely bandwidth-bound — it also spends real time in compute and thread synchronisation. This figure is an assumption, not a fitted constant: no CPU measurement is in the calibration set.
CPU-only inference: no GPU is assumed. A GPU of any size will beat these figures substantially.
Throughput is a memory-bandwidth roofline estimate, not a measurement. It is labelled "estimated" everywhere it appears.
KV cache is computed from this model's published attention configuration.

FAQ