What fits in 8 GB of VRAM
Data updated . Last reviewed the same day, when this catalog snapshot was written.
Usable memory here is 7.6 GB, which is 95% of 8 GB. 15 of 49 catalog models fit at Q4, Q8, or FP16. 9 catalog GPUs have this much VRAM.
Models
Sorted by the highest quant that fits, then by headroom. The tok/s estimate uses GeForce RTX 5060 bandwidth only.
| Model | Best quant | Estimate | Headroom | Est. tok/s |
|---|---|---|---|---|
| Llama 3.2 1B Instruct | FP16 | 3.0 GB | 4.6 GB | 181 tok/s |
| Llama 3.2 3B Instruct | Q8 | 3.4 GB | 4.2 GB | 131 tok/s |
| Gemma 3 4B | Q8 | 4.1 GB | 3.5 GB | 108 tok/s |
| DeepSeek V4 Flash | Q8 | 4.5 GB | 3.1 GB | 2633 tok/s |
| Phi-4 mini | Q8 | 5.2 GB | 2.4 GB | 110 tok/s |
| Phi-3.5 mini | Q8 | 7.3 GB | 0.3 GB | 110 tok/s |
| Llama 3.1 8B Instruct | Q4 | 4.9 GB | 2.7 GB | 91.0 tok/s |
| Gemma 4 E4B | Q4 | 5.1 GB | 2.5 GB | 90.0 tok/s |
| Qwen2.5 7B Instruct | Q4 | 5.2 GB | 2.4 GB | 95.7 tok/s |
| Mistral 7B Instruct v0.3 | Q4 | 5.4 GB | 2.2 GB | 102 tok/s |
| DeepSeek R1 Distill Llama 8B | Q4 | 6.0 GB | 1.6 GB | 91.0 tok/s |
| Qwen3 8B | Q4 | 6.2 GB | 1.4 GB | 89.1 tok/s |
| Ministral 3 8B | Q4 | 6.3 GB | 1.3 GB | 86.2 tok/s |
| Gemma 2 9B | Q4 | 6.8 GB | 0.8 GB | 65.5 tok/s |
| Gemma 3 12B | Q4 | 7.3 GB | 0.3 GB | 61.4 tok/s |
Next misses at Q4: Gemma 4 12B (7.7 GB), Mistral Nemo 12B (8.8 GB), Ministral 3 14B (9.6 GB), Qwen3 14B (10.3 GB), Phi-4 (10.6 GB).
GPUs at 8 GB
- GeForce RTX 5050
8.0 GB vram · $249
- Radeon RX 7600
8.0 GB vram · $269
- GeForce RTX 5060
8.0 GB vram · $299
- Radeon RX 9060 XT 8GB
8.0 GB vram · $299
- GeForce RTX 4060
8.0 GB vram · $299
- GeForce RTX 5060 Ti 8GB
8.0 GB vram · $379
- GeForce RTX 3060 Ti
8.0 GB vram · $399
- GeForce RTX 4060 Ti 8GB
8.0 GB vram · $399
- GeForce RTX 3070
8.0 GB vram · $499
Questions
- How many catalog models fit in 8 GB of VRAM?
- 15 of 49 fit at Q4 or better inside 7.6 GB usable (95% of 8 GB).
- Which catalog GPUs have 8 GB of VRAM?
- GeForce RTX 5050, Radeon RX 7600, GeForce RTX 5060, Radeon RX 9060 XT 8GB, GeForce RTX 4060, GeForce RTX 5060 Ti 8GB, GeForce RTX 3060 Ti, GeForce RTX 4060 Ti 8GB, GeForce RTX 3070.
- Is the tokens/sec column a benchmark?
- No. It divides the GeForce RTX 5060's 448 GB/s by active weight size. Other 8 GB cards are slower or faster than that.