What fits in 192 GB of VRAM
Data updated . Last reviewed the same day, when this catalog snapshot was written.
Usable memory here is 182 GB, which is 95% of 192 GB. 44 of 49 catalog models fit at Q4, Q8, or FP16. No GPU in this catalog ships with exactly 192 GB. The list is the memory budget, not a product roundup. 1 of the fits do not fit the 128 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 | 179 GB | — |
| Llama 3.2 3B Instruct | FP16 | 7.7 GB | 175 GB | — |
| Phi-3.5 mini | FP16 | 9.2 GB | 173 GB | — |
| Phi-4 mini | FP16 | 9.2 GB | 173 GB | — |
| Gemma 3 4B | FP16 | 10.3 GB | 172 GB | — |
| Mistral 7B Instruct v0.3 | FP16 | 17.4 GB | 165 GB | — |
| Qwen2.5 7B Instruct | FP16 | 18.3 GB | 164 GB | — |
| Gemma 4 E4B | FP16 | 19.2 GB | 163 GB | — |
| DeepSeek R1 Distill Llama 8B | FP16 | 19.3 GB | 163 GB | — |
| Llama 3.1 8B Instruct | FP16 | 19.3 GB | 163 GB | — |
| Qwen3 8B | FP16 | 19.7 GB | 163 GB | — |
| Ministral 3 8B | FP16 | 21.4 GB | 161 GB | — |
| Gemma 2 9B | FP16 | 22.2 GB | 160 GB | — |
| Gemma 4 12B | FP16 | 28.7 GB | 154 GB | — |
| Gemma 3 12B | FP16 | 29.2 GB | 153 GB | — |
| Mistral Nemo 12B | FP16 | 29.4 GB | 153 GB | — |
| Ministral 3 14B | FP16 | 33.5 GB | 149 GB | — |
| Phi-4 | FP16 | 35.2 GB | 147 GB | — |
| Qwen3 14B | FP16 | 35.4 GB | 147 GB | — |
| Qwen2.5 14B Instruct | FP16 | 35.4 GB | 147 GB | — |
| gpt-oss-20b | FP16 | 50.2 GB | 132 GB | — |
| Mistral Small 3.2 24B | FP16 | 57.6 GB | 125 GB | — |
| Gemma 4 26B-A4B | FP16 | 61.9 GB | 120 GB | — |
| Gemma 2 27B | FP16 | 65.3 GB | 117 GB | — |
| Gemma 3 27B | FP16 | 65.8 GB | 117 GB | — |
| Qwen3.6 27B | FP16 | 66.7 GB | 116 GB | — |
| Qwen3.8 27B | FP16 | 66.7 GB | 116 GB | — |
| Qwen3 30B-A3B | FP16 | 73.3 GB | 109 GB | — |
| GLM-4.7-Flash | FP16 | 74.9 GB | 107 GB | — |
| Gemma 4 31B | FP16 | 75.1 GB | 107 GB | — |
| Qwen3 32B | FP16 | 78.6 GB | 104 GB | — |
| DeepSeek R1 Distill Qwen 32B | FP16 | 78.6 GB | 104 GB | — |
| Qwen2.5 32B Instruct | FP16 | 78.6 GB | 104 GB | — |
| Qwen3.6 35B-A3B | FP16 | 86.3 GB | 96.1 GB | — |
| Mixtral 8x7B | FP16 | 112 GB | 70.3 GB | — |
| DeepSeek R1 Distill Llama 70B | FP16 | 169 GB | 13.1 GB | — |
| Llama 3.3 70B Instruct | FP16 | 169 GB | 13.1 GB | — |
| Qwen2.5 72B Instruct | FP16 | 174 GB | 7.9 GB | — |
| DeepSeek V4 Flash | Q8 | 4.5 GB | 178 GB | — |
| gpt-oss-120b | Q8 | 63.7 GB | 119 GB | — |
| Llama 4 Scout | Q8 | 115 GB | 67.9 GB | — |
| Mistral Small 4 | Q8 | 127 GB | 55.7 GB | — |
| Mixtral 8x22B | Q8 | 151 GB | 31.1 GB | — |
| Qwen3 235B-A22B | Q4 | 144 GB | 38.7 GB | — |
Next misses at Q4: Llama 4 Maverick (243 GB), Llama 3.1 405B Instruct (244 GB), DeepSeek R1 (405 GB), Mistral Large 3 (407 GB), DeepSeek V4 Pro (959 GB).
GPUs at 192 GB
No catalog GPU has exactly 192 GB of VRAM.
Questions
- How many catalog models fit in 192 GB of VRAM?
- 44 of 49 fit at Q4 or better inside 182 GB usable (95% of 192 GB).
- Which catalog GPUs have 192 GB of VRAM?
- None. No GeForce or Radeon row in this catalog is exactly 192 GB.
- Is the tokens/sec column a benchmark?
- No column is filled in, because no catalog GPU has 192 GB to borrow a bandwidth from.