Can granite 3.2 8b instruct run on the RTX 3070?
Yes. At 8K context the best fit is Q4_K_M, using about 6.3 GB of the 7.6 GB usable on the RTX 3070. Estimated speed ceiling: ~91 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 4.0 GB Fits | 4.6 GB Fits | 8.4 GB No | 23.4 GB No | 144 |
| 3-bit (Q3_K_M) • | 4.8 GB Fits | 5.5 GB Fits | 9.2 GB No | 24.2 GB No | 112 |
| 4-bit (Q4_K_M) • | 5.7 GB Fits | 6.3 GB Fits | 10.1 GB No | 25.1 GB No | 91 |
| 5-bit (Q5_K_M) • | 6.5 GB Fits | 7.1 GB Tight | 10.9 GB No | 25.9 GB No | 77 |
| 6-bit (Q6_K) • | 7.4 GB Tight | 8.0 GB No | 11.7 GB No | 26.7 GB No | — |
| 8-bit (Q8_0) • | 9.2 GB No | 9.8 GB No | 13.6 GB No | 28.6 GB No | — |
| 16-bit (F16/BF16) | 16.3 GB No | 17.0 GB No | 20.7 GB No | 35.7 GB No | — |
• real GGUF file size from bartowski/ibm-granite_granite-3.2-8b-instruct-GGUF. Speed is a bandwidth ceiling (448 GB/s), not a benchmark. Methodology.
More
granite 3.2 8b instruct on every GPU and Mac · Every model that runs on the RTX 3070 · Other models that fit 8 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (ibm-granite/granite-3.2-8b-instruct), NVIDIA specs. Data as of 2026-11-17.