Can I Run EXAONE 3.5 on 24 GB system RAM?
作者： Jakub Rusinowski · 最后更新： 2026年2月10日
Yes — comfortably
Yes, comfortably — EXAONE 3.5 7.8B at Q8_0 needs about 10.2 GB of the 19.2 GB usable on 24 GB system RAM, leaving ~9 GB spare and running at ~7.5 tok/s (estimated), with room for about 32,768 tokens of context.
Confidence: high · Recommended quantization: Q8_0 · Estimated speed: ~7.5 tok/s
See what else this hardware can run →
或在 Vast.ai 比较，低至 $0.35/小时 (typical low · varies)
作为亚马逊联盟成员，我们从符合条件的购买中获得收入。云 GPU 链接为推荐链接——我们可能获得佣金，您无需额外付费。
24 GB system RAM — what it gives a model
| Usable memory for models | 19.2 GB |
| Memory bandwidth | 90 GB/s |
EXAONE 3.5 on 24 GB system RAM: memory by quantization
| Quant | Memory needed | Fits 19.2 GB? | Max context | Est. speed | Download |
|---|---|---|---|---|---|
| F16 | 17.5 GB | ✓ Yes | 16K | ~4.1 tok/s | 15.6 GB |
| Q8_0 | 10.2 GB | ✓ Yes | 32K | ~7.5 tok/s | 8.3 GB |
| Q6_K | 8.3 GB | ✓ Yes | 32K | ~9.5 tok/s | 6.4 GB |
| Q5_K_M | 7.4 GB | ✓ Yes | 32K | ~10.8 tok/s | 5.5 GB |
| Q4_K_M | 6.6 GB | ✓ Yes | 32K | ~12.5 tok/s | 4.7 GB |
| Q3_K_M | 5.2 GB | ✓ Yes | 32K | ~16.7 tok/s | 3.3 GB |
| Q2_K | 4.4 GB | ✓ Yes | 32K | ~20.7 tok/s | 2.6 GB |
Which EXAONE 3.5 sizes fit
| Variant | Needs at Q4_K_M | Fits? | Est. speed |
|---|---|---|---|
| EXAONE 3.5 32B | 22.3 GB | ✗ Too large | — |
| EXAONE 3.5 7.8B | 6.6 GB | ✓ Fits | ~12.5 tok/s |
| EXAONE 3.5 2.4B | 2.9 GB | ✓ Fits | ~35.1 tok/s |
What to watch out for
- 1 larger variant of EXAONE 3.5 does 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.
- 19.2 GB of the 24 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
Can I run EXAONE 3.5 on 24 GB system RAM?
Yes, comfortably — EXAONE 3.5 7.8B at Q8_0 needs about 10.2 GB of the 19.2 GB usable on 24 GB system RAM, leaving ~9 GB spare and running at ~7.5 tok/s (estimated), with room for about 32,768 tokens of context.
Which quantization of EXAONE 3.5 should I use on 24 GB system RAM?
Q8_0 — it needs about 10.2 GB of the 19.2 GB available, downloads as roughly 8.3 GB, and runs at an estimated 7.5 tokens/sec with up to 32K of context.
What limits EXAONE 3.5 on 24 GB system RAM?
Nothing binding — the model fits with headroom and generates at a usable speed on this hardware.
Which runtime should I use?
llama.cpp (CPU build) or Ollama — both run without a GPU
Other RAM Capacities
Other Models on 24 GB system RAM
- Falcon 3 on 24 GB system RAM
- Gemma 2 Family on 24 GB system RAM
- Gemma 3 on 24 GB system RAM
- Gemma 3n on 24 GB system RAM
- Gemma 4 on 24 GB system RAM
EXAONE 3.5 on GPUs
- EXAONE 3.5 on NVIDIA GeForce RTX 5090
- EXAONE 3.5 on NVIDIA GeForce RTX 5060 Ti 8GB
- EXAONE 3.5 on NVIDIA GeForce RTX 5060
- EXAONE 3.5 on NVIDIA GeForce RTX 4090
What This Model Is Good At
Model & Tools
← Can I Run It? | EXAONE 3.5 model page | Check your hardware