What fits in 16 GB of VRAM
Data updated . Last reviewed the same day, when this catalog snapshot was written.
Usable memory here is 15.2 GB, which is 95% of 16 GB. 22 of 49 catalog models fit at Q4, Q8, or FP16. 14 catalog GPUs have this much VRAM. 1 of the fits do not fit the 12 GB tier.
Models
Sorted by the highest quant that fits, then by headroom. The tok/s estimate uses GeForce RTX 5080 bandwidth only.
| Model | Best quant | Estimate | Headroom | Est. tok/s |
|---|---|---|---|---|
| Llama 3.2 1B Instruct | FP16 | 3.0 GB | 12.2 GB | 388 tok/s |
| Llama 3.2 3B Instruct | FP16 | 7.7 GB | 7.5 GB | 149 tok/s |
| Phi-3.5 mini | FP16 | 9.2 GB | 6.0 GB | 126 tok/s |
| Phi-4 mini | FP16 | 9.2 GB | 6.0 GB | 125 tok/s |
| Gemma 3 4B | FP16 | 10.3 GB | 4.9 GB | 112 tok/s |
| DeepSeek V4 Flash | Q8 | 4.5 GB | 10.7 GB | 5643 tok/s |
| Gemma 4 E4B | Q8 | 8.3 GB | 6.9 GB | 117 tok/s |
| Qwen2.5 7B Instruct | Q8 | 8.6 GB | 6.6 GB | 119 tok/s |
| Mistral 7B Instruct v0.3 | Q8 | 8.8 GB | 6.4 GB | 125 tok/s |
| DeepSeek R1 Distill Llama 8B | Q8 | 9.6 GB | 5.6 GB | 112 tok/s |
| Llama 3.1 8B Instruct | Q8 | 9.6 GB | 5.6 GB | 120 tok/s |
| Qwen3 8B | Q8 | 9.9 GB | 5.3 GB | 110 tok/s |
| Ministral 3 8B | Q8 | 10.2 GB | 5.0 GB | 106 tok/s |
| Gemma 2 9B | Q8 | 10.7 GB | 4.5 GB | 89.8 tok/s |
| gpt-oss-20b | Q8 | 12.3 GB | 2.9 GB | 461 tok/s |
| Gemma 3 12B | Q8 | 12.5 GB | 2.7 GB | 76.7 tok/s |
| Gemma 4 12B | Q8 | 13.2 GB | 2.0 GB | 75.8 tok/s |
| Mistral Nemo 12B | Q8 | 14.4 GB | 0.8 GB | 73.7 tok/s |
| Ministral 3 14B | Q4 | 9.6 GB | 5.6 GB | 117 tok/s |
| Qwen3 14B | Q4 | 10.3 GB | 4.9 GB | 107 tok/s |
| Phi-4 | Q4 | 10.6 GB | 4.6 GB | 108 tok/s |
| Qwen2.5 14B Instruct | Q4 | 10.6 GB | 4.6 GB | 107 tok/s |
Next misses at Q4: Mistral Small 3.2 24B (15.7 GB), Gemma 3 27B (16.5 GB), Gemma 2 27B (16.6 GB), Qwen3.8 27B (17.0 GB), Gemma 4 26B-A4B (17.3 GB).
GPUs at 16 GB
- Radeon RX 7600 XT
16.0 GB vram · $329
- Radeon RX 9060 XT 16GB
16.0 GB vram · $349
- GeForce RTX 5060 Ti 16GB
16.0 GB vram · $429
- Radeon RX 7800 XT
16.0 GB vram · $499
- GeForce RTX 4060 Ti 16GB
16.0 GB vram · $499
- Radeon RX 9070
16.0 GB vram · $549
- Radeon RX 7900 GRE
16.0 GB vram · $549
- Radeon RX 9070 XT
16.0 GB vram · $599
- Radeon RX 6800 XT
16.0 GB vram · $649
- GeForce RTX 5070 Ti
16.0 GB vram · $749
- GeForce RTX 4070 Ti Super
16.0 GB vram · $799
- GeForce RTX 5080
16.0 GB vram · $999
- GeForce RTX 4080 Super
16.0 GB vram · $999
- GeForce RTX 4080
16.0 GB vram · $1,199
Questions
- How many catalog models fit in 16 GB of VRAM?
- 22 of 49 fit at Q4 or better inside 15.2 GB usable (95% of 16 GB).
- Which catalog GPUs have 16 GB of VRAM?
- Radeon RX 7600 XT, Radeon RX 9060 XT 16GB, GeForce RTX 5060 Ti 16GB, Radeon RX 7800 XT, GeForce RTX 4060 Ti 16GB, Radeon RX 9070, Radeon RX 7900 GRE, Radeon RX 9070 XT, Radeon RX 6800 XT, GeForce RTX 5070 Ti, GeForce RTX 4070 Ti Super, GeForce RTX 5080, GeForce RTX 4080 Super, GeForce RTX 4080.
- Is the tokens/sec column a benchmark?
- No. It divides the GeForce RTX 5080's 960 GB/s by active weight size. Other 16 GB cards are slower or faster than that.