Skip to content
localllm
日本語

Find by VRAM / unified memory

Local LLMs that run in 8 GB of VRAM

Of the 1,276 models listed on this site, 584 have a quantization whose estimated memory fits in 8 GB at an 8,192-token context. For each model we show the largest (most precise) quantization that fits.

Data as of

Models that fit
584
Largest model that fits
Qwen3.8-Flash-Next-FP4-Local-Strata 51.2B
8 GB devices (listed here)
25

Sorted by Hugging Face downloads (last 30 days, summed across sources). Not a performance ranking.

See the full download ranking →
  1. 1 Qwen3.8-27B 27.3B UD-IQ1_M 8.47 GB
  2. 2 Ornith-1.5-9B 9.2B Q5_K_M 8.22 GB
  3. 3 gemma-4-12B-it 11.9B IQ4_XS 8.49 GB
  4. 4 Qwen3.5-9B 9B Q5_K_M 8.48 GB
  5. 5 MOSS-Music-8B-Instruct Q4_K_M 7.24 GB
  6. 6 Ornith-1.0-9B 9B Q5_K_M 7.86 GB
  7. 7 Qwen3-VL-8B-Instruct 8.2B UD-Q5_K_XL 7.94 GB
  8. 8 Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP 9B Q4_K_M 8.35 GB
  9. 9 Qwen3.5-4B 4.2B UD-Q8_K_XL 7.32 GB
  10. 10 Ornith-1.5-9B-uncensored 9B Q5_K_M 7.83 GB

Estimated Memory is estimated at 8,192 tokens with an f16 KV cache.

Third-party evaluations of models that run in 8 GB

Top values from other organizations for the original models before quantization (not this site's rating).

Human preference (LMArena)

See all →
  1. 1.Qwen3.8-27B 1,433
  2. 2.gemma-3-12b-it 1,342
  3. 3.gemma-3n-E4B-it 1,317
  4. 4.gemma-3-4b-it 1,303
  5. 5.granite-4.2-8b 1,290

Source: LMArena (CC BY 4.0).

Capabilities (Epoch AI)

See all →
  1. 1.Qwen3.8-27B 152.7
  2. 2.Qwen3.5-9B 139.4
  3. 3.Qwen3-14B 136.2
  4. 4.Qwen3-8B 132.2
  5. 5.phi-4 129.1

Source: Epoch Capabilities Index (Epoch AI) (CC BY 4.0).

Devices with 8 GB of VRAM or unified memory

Choose a device to see the models that run on it and estimated generation speed.

All models that fit in 8 GB of VRAM

One entry per model, up to 30, sorted by popularity. For each model we show the largest (most precise) quantization that fits in this memory with a context of at least 8K tokens. The context length is the longest one that quantization fits (for models whose maximum is under 8K, such as embedding models, that maximum). Models that are not for chat or text generation are labelled with their use. "Left" is the memory minus the estimate; under 5% of the memory is marked "tight" (not a guarantee that it fits).

Frequently asked questions

Which local LLMs run in 8 GB of VRAM?
On this site, 584 models have a quantization that fits in 8 GB at 8,192 tokens (estimated). Popular ones include Qwen3.8-27B, Ornith-1.5-9B, gemma-4-12B-it, Qwen3.5-9B, MOSS-Music-8B-Instruct.
What is the largest model that runs in 8 GB of VRAM?
Among the models listed here, Qwen3.8-Flash-Next-FP4-Local-Strata (about 51.2B). With BF16 it needs an estimated 2.66 GB at 8,192 tokens. Smaller quantizations lower quality.
Is it the same for a Mac with 8 GB of unified memory?
macOS limits how much memory the GPU can use, so not all 8 GB is available to the model. On a Mac, leave headroom over the required memory.
What happens with a longer context?
The KV cache grows, so the required memory grows too. Each model page shows estimated memory for each context length.