What fits in 64 GB of VRAM
Data updated . Last reviewed the same day, when this catalog snapshot was written.
Usable memory here is 60.8 GB, which is 95% of 64 GB. 39 of 49 catalog models fit at Q4, Q8, or FP16. No GPU in this catalog ships with exactly 64 GB. The list is the memory budget, not a product roundup. 1 of the fits do not fit the 48 GB tier.
Models
Sorted by the highest quant that fits, then by headroom. Tok/s is blank because this tier has no catalog GPU.
| Model | Best quant | Estimate | Headroom | Est. tok/s |
|---|---|---|---|---|
| Llama 3.2 1B Instruct | FP16 | 3.0 GB | 57.8 GB | — |
| Llama 3.2 3B Instruct | FP16 | 7.7 GB | 53.1 GB | — |
| Phi-3.5 mini | FP16 | 9.2 GB | 51.6 GB | — |
| Phi-4 mini | FP16 | 9.2 GB | 51.6 GB | — |
| Gemma 3 4B | FP16 | 10.3 GB | 50.5 GB | — |
| Mistral 7B Instruct v0.3 | FP16 | 17.4 GB | 43.4 GB | — |
| Qwen2.5 7B Instruct | FP16 | 18.3 GB | 42.5 GB | — |
| Gemma 4 E4B | FP16 | 19.2 GB | 41.6 GB | — |
| DeepSeek R1 Distill Llama 8B | FP16 | 19.3 GB | 41.5 GB | — |
| Llama 3.1 8B Instruct | FP16 | 19.3 GB | 41.5 GB | — |
| Qwen3 8B | FP16 | 19.7 GB | 41.1 GB | — |
| Ministral 3 8B | FP16 | 21.4 GB | 39.4 GB | — |
| Gemma 2 9B | FP16 | 22.2 GB | 38.6 GB | — |
| Gemma 4 12B | FP16 | 28.7 GB | 32.1 GB | — |
| Gemma 3 12B | FP16 | 29.2 GB | 31.6 GB | — |
| Mistral Nemo 12B | FP16 | 29.4 GB | 31.4 GB | — |
| Ministral 3 14B | FP16 | 33.5 GB | 27.3 GB | — |
| Phi-4 | FP16 | 35.2 GB | 25.6 GB | — |
| Qwen3 14B | FP16 | 35.4 GB | 25.4 GB | — |
| Qwen2.5 14B Instruct | FP16 | 35.4 GB | 25.4 GB | — |
| gpt-oss-20b | FP16 | 50.2 GB | 10.6 GB | — |
| Mistral Small 3.2 24B | FP16 | 57.6 GB | 3.2 GB | — |
| DeepSeek V4 Flash | Q8 | 4.5 GB | 56.3 GB | — |
| Gemma 4 26B-A4B | Q8 | 27.2 GB | 33.6 GB | — |
| Gemma 3 27B | Q8 | 28.7 GB | 32.1 GB | — |
| Qwen3.6 27B | Q8 | 29.1 GB | 31.7 GB | — |
| Qwen3.8 27B | Q8 | 29.6 GB | 31.2 GB | — |
| Gemma 2 27B | Q8 | 30.0 GB | 30.8 GB | — |
| GLM-4.7-Flash | Q8 | 32.3 GB | 28.5 GB | — |
| Qwen3 30B-A3B | Q8 | 33.3 GB | 27.5 GB | — |
| Gemma 4 31B | Q8 | 34.0 GB | 26.8 GB | — |
| Qwen3 32B | Q8 | 37.0 GB | 23.8 GB | — |
| DeepSeek R1 Distill Qwen 32B | Q8 | 37.0 GB | 23.8 GB | — |
| Qwen2.5 32B Instruct | Q8 | 37.0 GB | 23.8 GB | — |
| Qwen3.6 35B-A3B | Q8 | 37.1 GB | 23.7 GB | — |
| Mixtral 8x7B | Q8 | 50.7 GB | 10.1 GB | — |
| Llama 3.3 70B Instruct | Q4 | 42.5 GB | 18.3 GB | — |
| DeepSeek R1 Distill Llama 70B | Q4 | 45.2 GB | 15.6 GB | — |
| Qwen2.5 72B Instruct | Q4 | 46.7 GB | 14.1 GB | — |
Next misses at Q4: gpt-oss-120b (63.1 GB), Llama 4 Scout (65.4 GB), Mistral Small 4 (74.0 GB), Mixtral 8x22B (87.5 GB), Qwen3 235B-A22B (144 GB).
GPUs at 64 GB
No catalog GPU has exactly 64 GB of VRAM.
Questions
- How many catalog models fit in 64 GB of VRAM?
- 39 of 49 fit at Q4 or better inside 60.8 GB usable (95% of 64 GB).
- Which catalog GPUs have 64 GB of VRAM?
- None. No GeForce or Radeon row in this catalog is exactly 64 GB.
- Is the tokens/sec column a benchmark?
- No column is filled in, because no catalog GPU has 64 GB to borrow a bandwidth from.