Can deepseek coder 6.7b base run on the RTX 3070?
Yes. At 8K context the best fit is Q2_K, using about 6.8 GB of the 7.6 GB usable on the RTX 3070. Estimated speed ceiling: ~177 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | Max tok/s (est., 8K) |
|---|---|---|---|
| 2-bit (Q2_K) • | 4.8 GB Fits | 6.8 GB Fits | 177 |
| 3-bit (Q3_K_M) • | 5.5 GB Fits | 7.5 GB Tight | 136 |
| 4-bit (Q4_K_M) • | 6.3 GB Fits | 8.3 GB No | — |
| 5-bit (Q5_K_M) • | 6.9 GB Tight | 8.9 GB No | — |
| 6-bit (Q6_K) • | 7.6 GB No | 9.6 GB No | — |
| 8-bit (Q8_0) • | 9.1 GB No | 11.1 GB No | — |
| 16-bit (F16/BF16) | 15.0 GB No | 17.0 GB No | — |
• real GGUF file size from bartowski/deepseek-coder-6.7B-kexer-GGUF. Speed is a bandwidth ceiling (448 GB/s), not a benchmark. Methodology.
More
deepseek coder 6.7b base 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 (deepseek-ai/deepseek-coder-6.7b-base), NVIDIA specs. Data as of 2026-11-17.